VOLUME TWO NASA SP-290 CASE FILE COPy TIII_ilINI] II)I_Sll;N .,ld AIDIDI, IIr,AI'ION NATIONAL AERONAUTICS AND SPACE ADMINISTRATION t,_,_
VOLUME TWO NASA SP-290
CASE FILECOPy
TIII_ilINI] II)I_Sll;N.,ld AIDIDI,IIr,AI'ION
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
t,_,_
NASA SP-290
TUleI|I§IBiDII]KIq;m.n.! AIDIDI,Iq',ATIII/
VOLUME TWO
Edited by Arthur J. GlassmanLewis Research Center
Scientific and Technical ln/ormation O_ce 1973
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
Washington, D.C.
For sale by the Superintendent of Documents
U.S. Government Printing Office, Washington, I).C. 20402Price $1.80 Stock Number 3300-00535
Library of Congress Catalog Card N_lmber 79-185105
PREFACE
NASA has an interest in turbines related primarily t<) ,wr<mautics and
space applications. Airbreathing turbine engines pr()vide jet and turb()-
shaft propulsion, as well as auxiliary power for aircraft, l)r()pellant -
driven turbines provide rocket propulsion and auxiliary p()wer for space-
craft. Ch)sed-cycle turbine engines using inert gases, org'mie fluids, and
metal fluids have been studied f()r providing long-durati()n ele('tric power
for spacecraft. Other applications of current interest for turbine engines
include land-vehicle (cars, trucks, buses, trains, etc.) 1)ropulsi(m powerand ground-based electrical p_)wer.
In view of the turbine-system interest and efforts at Lewis Research
Center, a course entitled "Turbine Design and AI)plicati_n" was l)re-sented during 1968-69 as part of the In-He)use Graduate Study l'r_)gram.
The course was somewhat revised and again presented in 1972-73. Various
aspects of turbine technology were covered including thermodynamic and
fluid-dynamic concepts, fundamental turbine concepts, velocity dia-
grams, losses, blade aerodynamic design, blade cooling, mechanical de-sign, operation, and performance.
The notes written and used for the course have been revised and edited
for pub]ication. Such a publication can serve as a foundation for an intro-
ductory turbine ct)urse, a means for self-study, or a reference f_r selected
topics. The first volume presented the material covering therm_dynamic
and fluid-dynamic concepts, fundamental turbine concepts, and velocity
diagram design. This second w)lume presents the material related tt)b]ade aerodynamic design and turbine energy losses.
Any consistent set of units will satisfy the equations presented. Two
commonly used consistent sets of units and constant values are given
after the symbol definitions. These are the SI units and the U.S. cus-
tomary units. A sing]e set of equations covers both sets ()f units by in-
cluding all constants required for the U.S. customary units and defining
as unity those not required for the SI units.ARTHUR J. GLASSMAN
o..lU
CONTENTS
CIIAPTICII. PAGI'_
I'REI,'A CE ........ iii
4 BL,.II)E DICSI(/.V hy W,ml<,r L. St<'w:_rt 'm(l Arthttr ,l.
(ll:l,_,_m:m ....... 1
S()LII)ITY . . . _ 2P, I,A 1)E- PllOF I I,t,_' I)ESI(;N ....... lS
IIE FI,_tlI+iNCES ............ 24
SYMBOLS ............. 27)
5 CItAN.VEL FLOII" .I.\ AL) ,I,_ I).v Th('o<l()r<' l(:_(s:mis _ 27
STI_EAM- AN1) POTENTIAI,-FIrN('.'I'ION AN.kI,YSI,;S . _ ;_;0
V1,;I,()(_I'I'Y-(',I{ A 1)I I_;NT ANALYSIS ......... 47
I_Ft,'EI/EN(_ES .......... 54
SYMI_()I,S ........... 55
(; INT'ROI)I'CT'IO.\ _ 7'0 BOI',\I)AI¢)'-I,..I)'ER TIll'gORY 1)y
Willi_m_ ]). 3I<'N.dly ___
NATUI,_E OF t),()I'NI)At_.Y LAYEI,I. ......
57
57
I)I_;liIVATI()N OF I_OUNI)AIiY-I,AYI,]I{ 1.;(_I:kTIONS .... (10
SOI,IrTI()N OF I_,OlrNI)AI>_Y-I,AYEI{ Fi(_I'ATIONS ....... 72C()N(_LI'I )IN(; I{EMAI_KS .......... S_
I_I,;I,'I,'=I{t+',NCI,;S................................... SS
SYMBOI:S ............................ Ol
7 B()I_NI)A RY-I,A )'ER L(),_SE,_ I)y H('rman W. ])rust, ,lr .... (,)3
BOIrNI)AI{Y-LAYEli I)AI_AMETERS ................. 95
BLAI)E-II()W LOSS COEFFICIENTS ........... 10l
I_;LAI)E-ROW LOSS C}IARACTEt_.ISTICS ............ 117
Ill';FEI{ENCES ................................. 121sY M I_() LS .............................. 123
8 MI,qCELLAA'EOI',q LOSSES 1)3" l{ich_trd .1. I{_)('lk('_ .... 12,")
TI I)-(_LI,2AICANCE LOSS .................. 125
I)ISK-FI{I(?TION LOSS ............ 131
I'A t{TI AL-AI )M I,qSION LOSSES ...................... l:_S
INCII)ENCE LOSS ...................... 143
REI,'E I{F,NCES .............................. 146
SYMBOLS ................ 147
v
CHAPTER4
BladeDesign
By WarnerL.StewartandArthurJ. Glassman
Th(, &,sign ()f a turbine e.nsists of three major steps. The first, is the
det(,rminal ion of th(' overall requirements of flow, work, and speed. These
are usually estal)lished by the particular application. The second step is
the ew)lution of velocity diagrams consistent with the desired efficiency
and/or mlmber of stages. This was discussed in chapter 3. The third step
is th(, design of the blading that will produce the flow angles and velocities
required by the velocity diagrams. This step involves the determination of
the size, shape, and spacing of the blades.
This chapter covers seine of the more important, aspects of blade design.
The height of the blad(' is set by the overall requirements of flow, speed,and inh,t state c,nditions and ttu, selected velocity diagram, which dictates
the fluid state conditions through.ut the, turbine. The blade chord is
usually seleet('d )-h(, a minimum vahw ccmsistent with mechanical con-si(h'rat ions. Th(, rh(sd must be long enough to allow accurate fabrication
and assure structur'll inl(,gri(y during operation. The selection of t)lade
spacing, which can be expressed nr)ndimensionally as solidity (ratio of
chord to spaeing) or axial s()lidity (ratio of axial chord to spacing), in-
volw's many considerations that will be discussed in the first part of this
chapter. Blade profile design, which includes blade exit and inlet geome-tries as well as the eom_ec)ing surface profiles, is then discussed in the last
part of this chapter. Channel flow theory, which is the basis for theanal::tical procedures used to accomplish the profile design, is discussed
in the next ehat)ter.
TURBINE DESIGN AND APPLICATION
SOLIDITY
One of the important aspects of turbitm blading design is the selection
of the blade solidity, which is tho ratio of chord or axial chord to spacing.
A minimum value is usually d(,sir(,d from the standp()int ()t" reducing
weight, cooling flow, and cost. ttow(w(,r, chord r('duction is limited t)3'm('chanical eonsid(,rations, and iner(,ased spacing ('v('nlually r('sulls indecreased blad(, ('ffi(!i(,ncy due to s('t)arat('d flow. This s(,ction will conc(,rn
its(,lf with tit(, a(,r()(tynami(' factors "lff(,eting s.lidity s(,h,cli(m. Th(, dis-
cussion will includ(' t lw ('ff('ct ()f v,+)cit y (liagr'm/ro(luir(,m(,nt s on solidityand th(, r(,lali(m b(,t w(,(,n I)h_(l(_h)a(ting and solidity. Als()in('lud('d will t)(, a
descrit)ti.n ()t' advanc('d blading c()n('(.pts that al'_' l)('in_z stu(ti(,d for us(' to
suppr('ss separati()n 'rod th(,r(,1)y r(,duc(, th(, ])(,rmissibh, s()li(tit.v.
Effect of Velocity Diagrams on Solidity
Figur(, 4-1 shows a typical s(.t ()f l)la(h, inh,t and _,xit diagrams as w(qlas th. static-pr(,ssur(, distribution ar()und a I)I.M(,. Th(, v(,h,(.iti(,s in this
figur(' ar(' shown as al)s()lute v('h)cili(,s. Tlw dis(:ussi()n in this chapt(,r
pertains to r()l()r |)lad(, rows as w(,l] as t_) star ()r t)]a(h, r_,ws. Wh(,n r(,f(,rringto a r()t()r, w(, must us(, r('httiv(, rath(,r than :d)s.htt_' v('h)citi('s in the
(,quations and figur('s. Sine(, in this ('hapt(w w(_ ar(, ('.n('(,rn(,d with I)ladc
rows rath(,r than with stag(,s, th(, angle (.)nv(,nt i.n will di ff(,r slighl ly from
that used in pr(,vi()us ('hapt(ws. "1"t1(,('xit tangential-v(,lo(.ity (:()mpon(,nt
and flow anglo ar(. tak(,n as negativ(, values. Th(, inh,t wdu(,s ar(, positive
if the inlet and exit tang(,ntial-veh)(,ity compon(,nts ar(. in ol)posit(_ dir(,c-tions, and n(,gative if in the same dir(,ction.
If one consid(,rs th(, two-dim(,nsional flow through a passag(' of unit
height betw('(,n two blades, th(,n tit(, tang(,ntial for('(, ('x(,rt(,d by the
fluid as it fl(tws from blad(, inh,t (sut)s('ript l) t() (,xit (subscript 2) is
1I",, =- ._p21".,2 ( V,,,1-- V,, ,o) (4-1)
g
where
F_
g
P
"_,Tx
_Y u
tang(,ntial fore(,, N; lb
c()nv(,rsion constant, 1 ; 32.17 (ll)m) (ft)/(lbf) ( s.("-' )
blitdo spacing, m; ft
(h,nsity kg/n?; ll)/ft :_
axial compon(,nt of v(qocity, m/'s(,('; ft/s(,c,
tangontial compon(,nt of v('locity, m/see ; ft/s(w
This t'mg('nti'd f()rc. ('x(,rt(,(t by th(, fluid must I)(, tit(, sam(, qs th(, for('(,
due to th(, statie-pr(,ssuro distribution around the bla(h,, :ts was discuss(,d
in chapt(w 2. Th(' h)w(,r part of figure 4 1 sh()ws a typical static-pressure
2
Stations
Vx, ]
Pl
T
_rface 11
-_"_ i; Vx,2
-in
Axial distance
BLADE DESIGN
]:IGI;RI'; 4 I. Typical bltuh_-row velocity diagrams and surface static-pressuredist rilmli.n.
(list rilmtion around t lw hlad(, row as a funet ion of axial distance. The area
])('tw('on th(' two em'v(,s rol)r(,s(,nts tho total Mad(, forco acting on the
flow in lb. tangential dir('('lion. Thus,
_.VII(_F( '
Pp
P.
The axial solidity, _, is
L' (t)l".=c. (p.-- p.J dC
axi,d chord, m;ft
l)ressuro-surf-t(!(' static pr(,ssuro, N/m_-; lt)/ft 2
suet ion-surfae(, star ie pr(,ssur(,, N/m°-; lt)/ft -°
axi'd distance, m; ft
('.r
G z
(4 2)
(4-a)
TURBINE DESIGN AND APPLI(!ATION
Substituting equations (4-1) and (14-2) into equation (4-3) the, yields
pWx,2( V,, ,1- V,, ,_)_ - (4-4)
/o'g (pp--p,) d
At this point, we introduce two tangeiltial loading coetlicients (hal have
been used to relate the actual blade loading to an ideal blade loading. The
first is the widely used coeificient b_troduced by Zweifel (ref. 1 ). This
coefficient is based (m an ideal loading t httt assumes ( 1 ) the stal ic pressur(,
on the pressure surface to b(' c(mstarlt aIld e(lual t() th(, inlet total pressure
and (21) th(, st._t ic l)ressure on the stwt ion surfac(, to t)e ctmst'm! and ('(lu'ttto the exit static pressure. I, ('(luati(m f¢)rm,
(p,- p,_) d
p,'- p._4-._)
where
_z
])1 _
1)2
Zwcifel loading coefficient
inlet total pressure, N/re"; 1t). ft _
exit star ic I)ressure, N/m-_; ll)/ft 2
The secoild coefficient is similarly defined <'xc('pt that lhe assmned
constant static pressure (m lhe suction sm'fa('e is ('qual t() the minimum
value of static pressure (see fig. 4-1) (m that surfa('e. This h_ading coeffi-
cient can never exceed a value of 1, a_d for all practic,d purposes, it must
always be less than 1. The Zweifel c,,eiticie,lt, ()n 1he ()lh('r hand, can ex-
ceed a value of 1. In equation form, this sec(,_d ],)adintz (',wtficient ¢/is de-fined as
1 (2£ (p,-p,) d .r
¢_=-., - (4-6])1 -- ])s ,m in
where p ....... is the minimum static l)ressure (m the suction surface inN/m "_(>r Ib/ft'-'.
The vcl<)city components in terms of velocity and flow angle are ex-pressed as
V,, = V sin a (4-7)
and
1"._= I" cos a (4-8)
IH,AI)E l) IG'q (; N
vvher(,
|" fluid velocity, m/see; ft/see
a fluid flow anglo, (leg
Substitulin_equ'dions 14 5) or (4-6), (4-7), and 14 8) into equation
(4 -t) and using lho lrig.nomelrie relalion sin 23=2 sin a cos a yields
(K--I sin2a., 2,(/ (K-1)sin2a2 (4 9)(7, = .... =
\P_ -/) .......,,/ ¢' \ pl -- p_./ ¢_
where K is the ral i,) of t'mgent ial velocity eoml)Onont (l',, A) at the bladeinh'l, to that (1"._.) at the blade exit.
l)erirotion of i.compres.sibh_-flo., rclatiol_s.--llelations involving solidity,
v('loeity diagrams, and h)a(li)lg are usually evolvod t)y assuming incom-
pr('ssibh' flow _vith no loss. With this assumpti(m, density p is constant,and l_('rnoulli's (,(luati(m
1
P'= P+ 2y pI'2 (4-10)
can be us(,(t. Substituting equation (4 10) into e(luation (4-9) yields
(K-l) sin 232 (K--l) sin2a2,r,= (4 11)
4' \ IV" /
_v]>r_' V..... is the w,loeity m_ the suction surface where p=p,.,._..
l.et us now define a Stletion-surfa(.o diffusion parameter D, as
1).=- U_'"' (4-12)I"2 2
._l'my l)ar:tmet(,rs of this type have l)e('n used to ret)resent a measure ofthe (h,(:oh,ral ion of the flow on the suction surface. This deceleration is an
indi('ati,m ()f the sus('('l)libility of the flow on the blad(, to separate. Using
thisd(,finition (('(t. (4 12)) in equation (4-11) yields
{K-- 1 ) sin 232 (K- 1 ) sin 23:_, - = (4-13)
e/1)_ J/.
l';(luatio.. (4 13) shmvs lhat the solidity parameter a.ffD, or a_ff. is
(:(mslant f_>r ,,aeh l)arli(_ular v('h)<qty-diagram re(tuirement. Since loading
('oettiei(,nt ¢, which oannot ('x('eed a value of 1, does not vary greatly, ite'm be se(,n that d('cr('asing solidity rosults l)rimarily in increased suction-
TURBINE DESIGN AND APPLICATION
surface diffusion (higher D,), the conse(tu(,nc(, of which will be discussed
later in this chapter. The solidity paranwt(,r is l)l()ited against the tan-
gential velo(:ity ratio K for several values of exit flow angle in figure
4-2(a). A wlhw of K=0 r(,pr(,s,,iits a r(,actio,l })]'id,, with axial inlet, a
value of K=- 1 r(,presents an ilnt)uls(' |)lad(', and "t value of K <- 1
retir('s('nts a n(,gative reaction blade. Posilive values of K r(,t)r(,s(,nt inlet
and exit tangential v('locities in th(, sam(, dir(,cti,)n and are ,'neountered
primarily in the tip sections of rotor blad('s. As seen from ('(tuati(m (4-13),
solidity param(,t,,r is equal to zero for all exit angles f(_r K = 1. This rel)re-sents the case wh('re there is no turning ()f the flow. The solidity l)aram(,ter
increas('s with (t(,('r(,asing K values. Thus, if exc[,ssiv(, suction-surface
diffusion is to b(' avoided, solidity must increase as the velocity diagrams
move from r(,a('li()n toward impulse. It can tie seen thttt for any giv(,n
value of K, a maximum value of solidity t)'trameter is (_l)tained with an
exit angle of 45 °.
Equati(m (4-13) can be m(_(tifi(,d to a function of the inlet and exit
angles to yMd th(. equation derived in ref,'r('nc(' 1.
2 COS _2 .
_- sm (a,- a2) (4-14)_z COS _1
For brevity, this is expressed only in t(,rms of the coefficient _b,. Equation(4-14) shows that th(, solidity p:mmwl('r a_b_ can be expr('ss('d in t('rms
of the flow angh's only. Solidity l)aram('ter is l)lotl('d against exit flow
angl(' for several values of the inh,t flow angle in figur(' 4-2(t)). For a
given (,xii angl(', solidity i)aramet('r increases with increasing inlet angl(,.
In t h(, region of most interest (a,> 0°, a_.< -45°), solidity parameter for
each inlet angle d(,creas(,s with (h,er(,asing exit angle.
A third relation can b(' (,v()lved, this ,me in t(,rlns of 1)lade reaction R,
which was (h'fi,wd in eh'q)t(,r 2 as
_12
R _ 1 -- -- (4-15)| "22
Substituting ('(tuaiion (4-8) into ('(luati()n (4-15) yMds
R= a-( c°s a2"]2 (4-16)\cos all
for the two-dimensional, incompressible-flow case, where V:.,=I'_,2.
Substitution of equation (4-10) tia('k inlo e(lualion (4 14) then yMds
2a_=-- %/] -It' sin Aa (4-17)
_/%
wkere Aa is ai--a._.
BLADE ])I",SI(;N
3
_2
e,J
(a)
0.5
Exit-flow angle,
a2,de9
-45
-60 or -30
_/./....../- 75 or -15
1 I I J0 -.5 -1.0 -1.5
Tangential velocity ratio, K
.. 3 r _lnlet-flow angle,
Y,
o1/-_ J I 1-30 -45 -60 -15
Exit-flow angle, a2, deg
1-_)
ID
r-
E
BO
2
1
0-. 25
- Turning angle,A_deg
f 90,, ,,- 60 or 120
0 .25 .50 .75 1.(Y)Reaction, R
(a) Effect of tangential-velocity ratio and exit-flow anglc.
(t)) Effect of exit- and inlet-flow angles.
(c) Effect of reaction and turning angle.
FIovaE 4-2.--Effect of velocity di,_grams on solidity.
TURBINE DESIGN AND APPLICATION
Equatior_ 4-17) expresses the solidity parameter iu terms of blade
reaction and turning angle. The solidity parameter is plotted against
reaction f()r several values _)f turning angle in figure 4-2(c). It can be seen
that, as indicated previously, the solidity parameter decreases with in-
creasing reacti_n. The solidity parameter is a maximum for a turning
angle of 91)° and varies little with turning unless very high or very lowturnings arc used.
Radical t'¢,riatiol_.--(;hapter 3 discussed the radial variations in velocity
diagrams that must (_ccur in order t(_ satisfy both the varying blade speed
and radial e(tuilibrium. Since axial sr)lidity was shown to vary with
varying veh)city diagrams, there will be a radial variation in the desired
value nf axial s_)lidity. The nature nf this radial variation will be illustrated
by an example. ('(resider a single-stage turbine having axial inlet and exit
flows (zero inlet and exit swirls), c(_nstant axial velocities, a constant
hub-to-tip-radius ratio of 0.7, an impulse rotor hub with a starer-hub exit
flow angle of -70 °, and free-vortex swirl distribution. For this case, the
flow angles at the hub and tip and the c()rresponding solidity-parameter
values computed from e(luation (4-14) are shown in the following table:
Stator It.(,tor
Inlet Exit Solidity Inlet Exit Solidity
angle, angle, parameter, angle, angle, parameter,
(teg (teg rr_¢, (leg (leg a_/,_
Hub ' 0 - 70 0.64 54 - 54 1.90
Tip 0 - 62 . 83 - 2 - 63 . 79
Note again that the angle convention being used in this chapter is some-
what different from that of previous chapters. Herein, stator exit angles
are negative. Assume that the loading coefficient _b, is to be maintained
constant radially. This is a reasonably desirable condition, and the
assumption enables us to proportion solidity directly to the solidity
parameter.Let us now determine how the hub and tip values of solidity parameter
shown in the preceding table can be made physically consistent. The axial
solidity variation in an), blade row must be inversely proportional toradius (because blade spacing is directly proportional to radius) and
BLADE DESIGN
directly proportional to axial chord. For the stator, the axial solidity
parameter at the hub is 0.64. If axial chord were held constant, then the
corresponding tip value of the axial solidity parameter would be0.64X0.7=0.45, which is almost half of the desired value of 0.83. There-
fore, a considerable axial taper from tip to hub is often used so that the
axial chord can increase with radius and yield the higher so]idities desiredat the tip.
In the case of the rotor, the axial solidity parameter at the hub is 1.90.
If axial chord were held constant, then the corresponding tip value of the
axial solidity parameter would be 1.90×0.7= 1.33, which is still larger
than the desired value of 0.79. Therefore, axial taper from hub to tip is
often used in rotor blades so that axial chord can decrease with increasingradius and yield the lower solidifies desired at the tip. Taper from hub to
tip in the rotor is not ()lily aerodynamically desirable, but is also mechani-
cally desirable from the standpoint of reducing blade stress. To simplifyfabrication in many cases, especially for smaller turl)ines, axial taper is not
used, and there results a radial variation in loading coefficient. With the
axial solidity selected on the basis of the mean-section velocity diagrams,
this radial variation in loading coefficient in many cases, especially thosewhere the blading is not highly loaded, will not have a severe effect onturbine performance.
Effect of compressibility.--The term
1
2g P_V2_
t
in equation (4-9) reduces to 1/D, for incompressible flow conditions, as
shown by equation (4-13). For a compressible flow case having the same
loading coefficient ¢_as for incompressible flow, division of equation (4-9)by equation (4-13) yields
1___ r 2
2g p2_ _ D,(7 x
o'z,i,_c pl' -- p_,rnin
(4-18)
where _,_._ is the incompressible flow value as determined from an
equation such as (4-13), (4-14), or (4-17). By introducing the relations
between critical velocity ratio, density, and pressure (eqs. (1-3), (1-52),
(1-61), (1-63), and (1-64)) and using the definition of D, (eq. (4-12)),equation (4-18) is modified to
TURBINE DESIGN AND APPLI(%_TION
y+l_' I 1-y-_1_4-1(_'%_) 2V2]I"("-I) (IVI_)ZD,z
(4-19)
where
7 ratio of Sl)ecific heat at constant pressure to specific heat at
constant 'v()]un] o
V,., critical velocity, m/see; ft/sec
Then, t)y using binoxlfial expansion and by neglecting the secondary terms,
equation (4 1(.I) call be al)proximated as
O"x
- 1 ..... + (4-20)z_,,,_ y+l 2(y+l)
The appr(_ximation represented by e(luation (4-20) is quite good for
(1"/'1",_)2 values up to about 1. The solidity ratio affz_,,, is plotted against
suction-surfac(, diffusion param(,t(,r f,r several values of critical velocity
ratio in figure 4 3. The compressibility effect bcconws more pronounced
as D_ eith('r i,er(,ases or decreases from a value of 2. At D, = 2, there is no
compressibility ('fleet for any value of (I'/1%_)2. For D._ values of less than
2, the required solidity decreases with increasing values of (V/V_,)2.
1.50 --
1.25-
.ff 1.00
.75
.5O
Exit critical-velocityVatio,
IVcr)2
1.2
1.0.6
0
I I2 3
Suction-surface diffusion parameter, Ds
Fmum.: 4-3. -Effect of compressibility on axial solidity.
10
BLADE DESIGN
For D, values of more than 2, a region that is only of academic interest
because it is beyond the limits of good design practice, the solidity ratioincreases with increasing (V/Vcr)2. Experience has shown that D8 values
should be maintained below about 2 to avoid excessive losses.
Relation of Loss to Solidity
It is well recognized that the loading of a turbine blade or of a com-
pressor blade is an important function of both solidity and reaction.
Correlation of blade loss with a compressor diffusion parameter was
described in reference 2 and is used widely within the compressor field.This parameter includes two terms, one reflecting reaction and the second
reflecting turning and solidity. An analogous diffusion parameter was
evolved for the case of the turbine in reference 3, where an overall diffusion
parameter is defined as the ratio of the sum of the decelerations in kinetic
energy on the suction and pressure surfaces to the exit kinetic energy. If it
is assumed that the pressure surface minimum velocity is low enough toneglect (Vp._n = 0), then the overall diffusion parameter is defined as
D--- V_,,- V::+ V12V: 2 (4-21)
With the use of the definitions of D, (eq. (4-12)) and R (eq. (4-15)),equation (4-21) reduces to
As seen from equation (4-13),
D=D,-R (4-22)
¢_=@D. (4-23)
Substitution of equations (4-23) and (4-14) into equation (4-22) thenyields
2 cosa2 .D= sin Aa-R (4-24)
ffx_ cos _1
This relation is like that for compressors, with the two terms involvingreaction and solidity.
Attempts have been made to correlate turbine blade loss with both
overall (ref. 4) and suction-surface (ref. 5) diffusion parameters. A
definite trend of increasing loss with increasing diffusion was established,but complete correlation could not be obtained. Such a correlation of
blade loss with diffusion parameter alone would not be expected, sincedifferent values of reaction and solidity giving the same value of D do
not give the same loss.
Consider first the effect of reaction on loss, as shown qualitatively in
figure 4-4(a). As reaction is reduced from a relatively high value near
11
TURBINE DESIGN AND APPLICATION
unity, there occurs a gradual increase in blade loss. Further reductions inreaction to negative values cause the loss to increase rapidly. This varia-tion in loss with reaction is caused by the change in boundary-layer
characteristics (which are discussed in chapter 6) as the nature of the
flow varies from highly accelerating to diffusing. The negative reaction
regime, although desired in many applications, is usually avoided because
of the high loss encountered when conventional blading is used.
The effect of solidity on loss is indicated in figure 4-4(b). A minimum
loss occurs at some optimum solidity. As solidity increases, the amount of
frictional surface area per unit flow is increasing. As solidity is reduced,
on the other hand, the loss per unit surface area is increasing because ofthe increased surfacc diffusion required. A minimum loss occurs as a result
of these opposing factors. The value of the suction-surface diffusion
(
oReaction,R
o
(b)
Axial solidity, ox
(a) Reaction.
(b) Solidity.
Fmvnr: 4-4.--Loss trend with reaction aad solidity.
12
BLADE DESIGN
parameter corresponding to the optimum solidity is a function of many
factors such as Reynolds number, shape of suction surface velocity dis-
tribution, and rate of turning, hi general, as mentioned previously, values
not exceeding about 2.0 arc used.
Selection of Optimum Solidity
Both analytical and experimental attempts have been made to identify
optimum solidity. According to reference 1, minimum loss occurs when the
Zweifel loading coefficient ¢Jz is equal to 0.8. By using this value in equa-tion (4-14), optimum axial solidity can be determined as a function of
the blade-row inlet and exit flow angles, and this is plotted in figure 4-5 (a)
for a wide rang(; of angles. The dashed (long-short) curve represents the
locus of points for impulse blading.
In order to determine the optimum values in terms of actual solidity,
it is necessary to determine the stagger angle a,, because
_x
a_-- (4-25)COS _#
An analytical blade model was used in reference 6 to relate stagger angle
to the flow angles and the axial solidity. Thus, optimum values of actualsolidity were obtained as a function of inlet and exit angles, as shown in
figure 4-5(b). The authors of reference 6 compared an optimum soliditydetermined in this way with the data of reference 7, where efficiency was
measured with four different rotor solidities, as shown here in figure 4-6.
The solidity determined as optimum in reference 6 from a figure such as
figure 4-5(b) is seen to be quite close to that yielding maximum efficiencyfor this case.
Loss coefficients based on cascade data are presented in reference 8 as a
function of pitch/chord ratio (inverse of solidity) and exit angle for
reaction blades (al = 0) and impulse blades (al = - a2). These coefficients,
in relative terms, are replotted here in figure 4-7 against solidity for
various exit angles. These curves indicate the importance of selecting
optimum solidity. For the larger (more negative) values of exit angle, thecurves are rather flat in the region of minimum loss, and some deviation in
solidity from optimum does not cause any significant increase in loss. As
the exit angle gets smaller, the minimum loss region becomes more pro-
nounced and the loss penalties become more severe as solidity departs
from the optimum value. It must be recognized that curves such as those
of figure 4-7 are usually obtained by using a given blade shape and varying
the spacing. Thus, the blade shape and resultant velocity distributioncannot be optimized for each solidity, and the significance of such acorrelation is somewhat clouded.
13
TURBINE DESIGN AND APPLICATION
Inlet-flow\-- \ angle,
\ \ \ _ Typeof
_ _ _ _)_ bladerow
\ \ _ Accelerating_ "" _ \ .. \ ------ Decelerating
. "_ _ _u \ Impulse
__ 3 - 4o \ \\\\k \\
i'L_ (a)o/ I 1 I I I 1
_4I0
32_3
0-20
Inlet-flow- \ \ ang4e,
\ \ %\ \ deg
\ \\ N7o
- \\6o \ \\ \\ \ \
\-- 4o \ \ \
\ k \
-30 -40 -50 -_ -70 -80
Exit-flowangle, ap deg
(b)
(a) Axial solidity.
(b) Actual solidity.
FIGURE 4-5.--Effect of inlet and exit angles on optimum solidity. ZweifeI loading
coefficient _, = 0.8.
14
.90--
•_ .88._..9_
;"5,
.86.8
BLAI)E I)t'_SI(;N
Optimum solidity from Number• - _ of
blades
o 24o 32o" 64
I I I I I _1.2 1.6 2.0 2.4 2.8
Solidity, o
lrmtm_: ,I _i. Vari'ation .f efficien<'y with solidity for four turbines ¢,f refermwe T.
..,,t-
o
-- Reactionblades (oI =O) Exit-flowI0-------- Impulse blades (oI = -02) angle,
\\ %deg
8- \ _-70
6_ _ _-4u
-60
2 _ "_ -40
01 I I I I I1.0 1.5 2.0 2.5 3.0 3.5
Solidity, o
l:mtrm,: 4 7. l'iffe('t .f s<,lidity and exit :ingle .n hl:uh' l.ss (,<_efli('i('nt.
The optimum soliditi('s ot)tained from the cascade resttlls shown in
figure 4 7 are plott('d against exit angl(, in figure 4-_ "rod qr(, (,(mqmrod
with those obtain('d analyti('tflly and shown in fi_zure 4-5 (1>). It ix ot)vious
that agreement betwe('n tit(, (,xperim(,ntal "rod lhe atmlytir'd results is
not good for most exit-angle values. Although tim (,xp(,rintenta[ and the
analytical cttrves do cross ettch other for both the rea(!l ion '_1 = 0) ))lading
attd the impulse ((n = - e_o) hln(lin_, th(, indieah'd variations in opt illltll/l
solidity with exit angle arc just not sitnilar. All thai can t)(' said at this
time is that the analytical r(,sults inv<)lv(' mtmy assumt)tions, t}l(, (,xp(,ri-
mental r(,sulls pert'tin to <m(' parti('ul'n" ]_]a(h' t)r(_fih', 'rod ther(, qre many
factors that act t() (h,t(,rmin(' .ptimtml solidily in n matm('r lhat w(' (h)
15
TUtiBINI_] I)ESll;N AND API_I.Ii'ATI(IN
-- Basedon analytical results of fig. 4-_b)
------ Basedon experimental results of fig. 4-7
3 Inlet-fluw angle,of
..._....-----._7_. _ de9S LI
o I I 1 I I _2-30 -40 -50 -60 -70 -80
Exit-flow angle, e?, deg
["l_ilrt{]; I S. (!<mllmt'i>_m ,_l',q_th,mtn s,_lfiliii,..s.
not yet fully und,,rstl/nd. Aualytic.d r_,sttlls, sm'h ns _h_,s_, <,1' figure 4 5,
qre more fre{itwtdly usi'd to d_.ternline ()1)i imtull s(,lidii v 1hltll Ill'(' ('xperi-
mental I'('>_lllts, SlI{'}I _.IS [}I()SC SII, IWI! ill tiKur{' .1 7. ('llrr{q_t &'sign l)raclice
is t_) use _ v:thics of 0.9 t_l I.(), x, llicli is sli/hll3 }iivhl'r than the 0.g
]'(W()llllll(ql(liq{ in I'('[('I'(ql('(' 1.
Ultralow-Solidity Blading
In tit{, p'tst, thv lilnilati{m ici r_,du('ti,m.'- h, s,,liditv has been separalion
,w(!urring _>tl th<' sttcti_>n Slll'facl, of Ill(, llla,l+,. '1"<+:tdli,'v,' lmvcr solidilies,
s<ml_' l,iodiii<'ati<,r ill hhtdu c_lnceI)l liillSt i,!' uiilix<',l S/t,qi thai Sl'lmralion
is snpl_rcssed au<l ill_, ass(wiat,_t high h,>_,.s d_ Hid lJt'(qll'.
Tt> tl'{':lIlil('lll of tiw ])l)ltlld:lI'V ]a\(T ill l.h(, l'{'_i{m id' s(,l);traiion is OllC
apl:lr(>:.tc]l ill re(hlc,'_] ,,<,lidiiy. Sttch tl'l';tillifqlf s e,,uht tilt'hid{' I'('IIIOV]II_ t ]1("
J;,()tllid.:tr.v laver t,v Stlclil>ii, i,ilc,z'_iZiilX l lie IJlJtilidlil'y ]ayl,r })y lilov, in_, (it"
incrcq.shig; lhi, lttl'iJtlllqi('c (d' lhl, lt_,tliidarv hty('r t>y tl>(' ,li' lliil)li]_l.l_)i's ,",it
the I)hich,. ('c,rlain ill Ihesl, i,Otl('{,]>ts hav(, IJ(,(ql ext)l_Ji'cd with nlar<_inal
SllCe('sS. T',\o a]ll,l'llilte I)]ade ('{lli('('t)ls l tlai }laVe, l_erhal)S, ])(qt('I' l)otential
:i.I'(' l ii(' lalldClli itllt[ jel-llal)t>l:l@'s. "_\liich are ilht< ral!'d hi figure ,t-(t.
Sttidi_'s at)plyili K the l)Olllldary-lay(T tl'(,ltilll('lil C{illCellls as _'11 its the
Itlll'l'il,:tto lihtdc' c+ml"olits 1,) sl:itllr blades and i'(_I_1" 14ad(,s lit,, sUnllnttrized
in I'(,[(,l'l'llcl's !t ;tu(l l(i, respcclively. ('ascad{' t(,sts <d' h,w-s,,lidily l)]tt.in,
lanldt'lli, alia jl,t-lht 13llkuh,s ttl'(, i:lr(,s('ilt('<i i/i I'(,t'cr_.,liCCs 11 to 14. Tllrliin(_
lest resillls with h>xv-s<)liditv lalid_,nl and j(,l-fl,till l't_t_,ls all! prest,nted in
l'Cfl'l'tqll'('s 15 and Ill, r('spi'('tiv('ly.
Till, lalldl'lll t_la&' ()p(,r:tt_,s +ill llie princit)h' tll:d, :llllu>ugh _l tiitz>]l value
of suct il>li-suri'ac(' difl'usilm is utiliz<'d (t)('r]i:itis 2 ). lhe fr_>Iit flail is ter-
minatl,d :it ,:tltotil the plfint of sel)zir:tlilm. Tho rcnlainilig diffusion then
16
Tandem blades
],'x_;vi_l.: I !L
t/l,,',, D I,: l)l':_[_; N
Jet-flap blades
E
1.2
1.0'
.8
.61
.4
.2
-- 0 _I _fnf(4 percent flow)
Suction surface
_ Pressure surface __
I I I I I20 zlO 60 80 100
Axial chord, percent
17
TI'ItBINI_ I)I.]_V;N AND AI'PLI('AT[_)N
ar(Jund on(' such blade with th(. j_'l (d'f and on. With the' j('t on, th('re is no
l()n_z('r a requir('n,onl f(lr th(' v(+)citi('s ()n lh(' suet i()n and [Ir('sstlr(' 81|rf'Ic('s
t_) bt' (,qual ,it th(, blade trailing (,dg('. Th(' 1,)'tdin_ diagram now apilro't('h(_s
a r(wttm_zular shap(', xvilh th(, l(Jad (.(_(,tii(.i(,nl 4' m_,r(' cl(Js(,lv al)l)roa('hing
unils'..\Is(), th(' dift'usi,m (tn tit(, su(:li_m surfact' ix sut)slanlially r_'dtw('(t,
lhus sUill)r('ssi|_ t h_'t(,n(l('n('y t (_ s('l/arat('.B()t h t h(, j(q -flap-I dad(' and _lmt and('m-I )la(h' ('_tn('('l)t s _ff('r t h(_ })(Jr('n-
lial f_,r s_di(tily r(,du(gi(ms. Th(, j[q flap, h()xv(,v[w, will llr(_l)al)ly b(' eon-
si(l(,r(,d o_dy f(w at)lllic:tti_ms xv|wr(' a s(w_mdary :tit- fhlw is r_'(tuir('d for
oth(,r pm'p_s('s, such as 1,lad(' ('(l(ding.
BLADE-PROFILE DESIGN
Aft('r 1tl(' ldmh' <'hord l('nlzlh has ll<','n ._[d(,('l_'d and lh(, |)lad( _ spacing
(l(_l_,rmin_d from1 s_,li(til,," (_onsi(h,rati_lnS, the, lllad(' i_s_.li' must 1)(' (t(,sign('d.This inw)lv(,s d,,t(,rminati(m _tf lh(' inl(q and _'xil _(,om_qri_'s "rod tlw
eonn_'('li|lg surfa('(' t)r_di]('s. Tim inhg and (,xil parts _d' th(' Illa(t(' must bc
(t(,sign('d 1,_ w, wi(h' a sm(_)lh, (,[ti('i(,n( lr:msiti_tn b(qw('('n th(, blade(!]I,'./[IIH']:111(ttilt' ['l'_(_S[['('.[/tll. Th(' surfa('(' llr_lfih's r'()ml('('ting lit(' inl(q and
(,xil_ mu._( llr(wid_' _h(' r(,quir('d tlmv _urnin_ wil h minhnmn htss.
Exit
(',msid('ration of lh(' Ill:M(' _'xit s_'(qi(m in('ht(h's th(' lr'_iling ('dg_', tim
thr_)at, and lh(' sucti(m surfat'(_ t)_,l_v(_('n tim thro:tl and Ill(' trailing (,(tg('.
T_',/lh_9 _¢fffc. In lh(, d_'si_n of turllitws, it is xvis(, l(_ uliliz(' lit(, sm'ill(,.,.'tlr ilin_ ('(t_z(' u_,nsisl('ni wilh ml,chani('al el)nsitl(,rali_ms. ,,ks shown in
r(,l'(,r(,n('(' 17, 'm in('r_'as(' in Irailing-('(lg(' lhi('kn['ss _':m._'s an iner(,as(' inlit(' Ida(t(' l_)ss. 'l'hi._ (,ff(,ct is (tis('uss('d furlh('r as part _)t' tlw turl)in('-l_ss
(liscus._i_m in (!httl)l_'r 7. In :_d(tili(m, lr'61ing-('d_(' (hi(.kn(,s.'_ also has a
si,,z,'nifi(':mt (,l'f_,(:l ()n the, fhlw l)h)('l_ag_' in ih[, lda(l_' (,xil r(,_i(m.('()nsi(l('rali_m _i' lit(' bl_'k:t_(' (,ff_,('l will b(' mad(' with lh(' us(' (_f tigur_
4 1 l, _hieh sh,,ws _'xa|nlll(' Ida(l(' s(,(gi_lns wilh lh_' n¢)m_'n¢'latur(' us('d. A
n(_w _,xit-v(,l(wily diagram is ('(msiru('l('d ,'tt sl'tii_m 2_r, whi('h is bw'tt('d
jusl within lh(' Ida(t(' lrailin_z-('diZ(' r(,_ion. Th(. r(,du('(,(I :m'a (hw l(_ timtrailit,_-('(l_z(' ld(wkag(' r(,sulls in a higlwr v(,l(wily :_1 slati_m 2a than at
station 2, _vhi_'h is I(wat('d .iusl b(w(md th(, bla(t(' mdling-(,(t_(, r(,gion.
"l'h(, ('(lua_i(ms lhat have' 1)('('n us('d t_) (d)tain lhis "within-tlw-blad("'
(tia/ram :_1 2:_ inchM(' (_ons('rvaii_m ,)f tang_'niial m(mt(,n(um:
V,,,_,_= V,,o (4 2_i)
and eontinuily :
( ' )' V (4-27)_p .,2. 1 (pl'_),.,
18
_ow Station
]
]'II,A I) E ])I_]SIG N
];mum.:4 l]. Bladesevtion and n,mwnvlatur(,.
where/is the trailing-edge thickness, in meters m' fl'('t. Tim flow :m_h. a2,,
is determined from equations (4-26) and (4 27) by "tssumin_ t].. flow
between stations 2a and 2 to be either incompressible (since the eh:mgos
are usually snmll) or isentropie. The hi'Me must be designed to have "m
exit an_le of a.,.. in order to produce a w'locity-dingram nngle a2 at station
2 outside the bl'tde row.
'1'|1¢, 3[ach mmd_er at station 2t_ can also be determined fr_ml the
preceding equations and assumplions. Because the angle a.. is often large
(63 ° or _ro'itvr) and the flow _\htcll uund.,r at the l)lade exit (st'tti,,n 2)
is .f'ten specified to b. in the high subs, role region, tim lrailin_-.dg.
blocka_e can cause st at ion 2a to bee_ mw ch,d¢ed. I t is, t here/'(,re, imp,,rl ant
to detormitw whether choking inside lhe blade row will r,ceur such that
the design flow r'tt e cannot be obt'fined.
Throat.--Since, in _eneral, a t urbino blade row .perates :is a nozzle, wit h
the flmv 'tc0eleratin_ up t_ the t hro'lt, .r minimum nrea, the determination
of the throat _penin/z, o (see fig. 4-11 ) hee_mws a rather m'itic'd aspect of
the th'sign procedure. ()he technique used successfully t_ _ivo this dimen-
si.n nml<es use of the "insi(le-the-trailin_-edVe" voloeity diag'ram. If ,,he
assumes no (,h:mg. in flow condilions "rod :t straight suction Sltl'f'teO
19
TURI_INI] DESI(;N AND APPLIt'.kTIt)N
).,tw(,(,u the' thr,,_tt a,,d stali<)u '2., th('rl th(, _hr,,'._l dim(.nsi(m can be
i,})laillt,d l'ri)lll t}l.c v(,h)ciiy ditl_rttlll at statism 2,, tLv using I]W following
('qllat i()II :
"(, 'i/= - ('()s<_,, (4 28),_ ,_,' (!()s 0_'2
xxlt(,r(, o is fllt, lhroat ()p('niri_, it'_ ]iwl(,rs (it' f('(q.
If it is :i_.>lll)l('d lh'l( (h(' v<,l<)cilv ;i)l(l l()ss (h) il,)l .h:ttl_(' l)(,l_x('('n ill('t]It'()_t( ;iIl_[ tilt' "I'I'('('-_tl'('alli" _tati<)n 2, lli('n
o
= ,'()s,_,: (t ')()),%.
k\'h(')_ '(hi-_ nlo)h()(l is ux('d. (lw (.ff('ct ,)fi)':d]i)la-(,d,.¢,. ihi<'kn('ss chtui_('s
(h(' an_l(' (,f ;h(. (}i)',):t( l)()sili,,n hut il,,l it_ l(.n_lh, l'l<)ih nwl]i()(Is ((,qs.(I '.2Si :rod I 2!),! ,.,iv<, ,..ilifi]'ir lhi'<,_i) <limc')isi(ms. I t('f('r('n('(' S COml)ar(.s
li_(.;tsur.([ (.×i)-fh)w :li)_l('>_ x_ith lh()._(, I)r('(li(')<'d l)v. ('(lti_i( i(m I:-I _.)).9(This
c',)mp;iri..)t_ i).li(':)t('._ ('l()s(' :i_)'('<'ni('t_l :it ('xil an.(.z,h's_.r('ai('r than 60 ° and
(h...'i:ll i(>)t._,)i' u)) (,) .-)°f()r (.×it _m_l<,s d,)w)l l,)3,-) °. ThS. <h,vii_(i<m e()uld be([u(, l<)]<)%_(,r .-,,li<li'_i('_ ;t,.. %\(,II ti_< htr,o('r _l':i(]i(,lltS I}itiI '.;())11(I()c('tll" tl('r()_s
Th<, )hi',,:ll-<>l,('ilin_ (lini(,ilsi<>)i "t,<d(q('rniiiwd fr,,m ('(luali())i (:4 '.2S)
()r tl 2!)_ ;tl)l)li,"_ i,, lh(' (':is(, _h(,r(, th(' l)l:.l(,-r<,x_ ,,xi( ll()\v is siibs()nie.
Ii' ih(' fl,,_ _i![liil lh_' l>bld(, r()w (.xliaii(Is t,);_ _lil>('i's(mi(' v(.locily, th(,)i
(his (',,Hq!_i)<,,i ([11'():it (lili.,)i_.i, >)inlu'_i 1,(' m<>difi(,(l i,, :i('('<)un( f()r <'×l)tltlsi()ii
fr<)))_ (he, --<,Ili..,_).liti,m :li ill<, (llr,>:t( 1(, (If(, _Ul)('i's(mi(' c())Idili()n a) (h('.
('xit. V,>r ('xit .\[:_('h )iml]l)(,rs _)'<,:)t(.r lll;IH ;d),)Ul I.:_, ih(' ('h(Jl<.iw_ s('(qi()n
_l}lD)itt } 1111l"-1 1)_' I( )('ttl ('(l })tlt'f; Wit}lill |}l('('}lt/llll('] ",tl('}l llltt) it ('OIlV('F_Ollf-
(liv('r_('Ni l)ti'.'-:t_,' i.', (,l)l:li)wd. l"()r ],)xx ".lII),')'s(,ili(' .\l',t('h mlml)(,rs (up to,
i)t,rhal)s, 1.3:,. i) ha>. l)<,(,)i f()Imd thal saIi.4tiv)(,rv l)(.rf<)rm'ulco can b(,a<.hi(.v(.d ir lh<, )}li.<,;l) 5< ..<till l<,(.at,.d al ih(' ('xi) ,,1")h<' ('h'lllli('l, H]I(I tilt'
•t, hliii,,ii:tl fl,>_ ('xl):ii,>i,,il ,,_','tlr: (l_)xvw-)r(':lln rr,,l,_ (}i_, (hr,)'l(. ll, tlff.<
(':is('. ih(' r(,(lilir(,d .h:uiu(,l (,xi( _li,li,,)l..i,))_ . _(>_II<I I)(, ('()mlm)('d l)\" lh('
1',,ll<)x_ilia ('(ill:it i_ >ll;
\vh(q.(,
tlir,):tt +>l)<'liiil_ (',)iillml('(l fr,)lil ('(lltati¢)H
_lll)('i"<( >r_i,'V<'I+.'it y, ill ; j'l'
[[t)_, _, tll'(':l t'()l' S!IIiI<' Itf)\V, 111'2; l'i 2
11')_ "il'(':i t'(>i' Stll)('l',",()ili(' ft()w, lii2; i'i _
,:.I 2S) ()r (4-'2!)) for
2O
l';I, '_ I)I_ I)I+_I(;N
1.0
I I I I.0 1.1 1.2 1.3 1.4
Mach number
l.'i+;I ](i ] I _'. ',,';tt.i;tti_m irl fl(:,v ar(_:l ,Ail}l <ql])(*l'_lllit. ttl_','_ _|:tl'}[ IJqlltt[)Pl'.
This area ¢',a'r,'¢'li<m, \',ith :tssum<'d is+'ntropic flmv lJ<'t\v_,,,n lhroat _u)d
exil, is shown in figttr+' 4 12.
.",',eli,, ._,_f,e+' d+,+v,._:cc+,, .i',,,+ H_co,t. The, s_.l+.+.ti+,ll +>f the' type' of_(IFI'II('(' I)('I\'CI'('II Ill(' |}11"11;)1 UIl{] trnilin_ +'(]_+'¢)11 t}l(' Stl(qi()II ,_)H'f;/('(' l'OlIS{
t7+, m:l(h, from stt+'h <'+msi&.rati,ms :ts slrtt('tura] ir)l_'_zz'il', in th( +tt'ailiu_z-('dg_' r+'gi+m, 31'u4t Imml_+'r h'vl'l an<l assoeiat_,d Iossv.-, (h,sir_,d h,vvl _f
stt<'ti_m-surl'a_'+, <liffusi,m I 1).,), and tTI,qdv surl':_<'_' :tt'_':_ r_'sttltin_ fv_m'_ ILL{'&'sign.
A "st r:ti_ht lmel,:" <h'si_t_ is us_'d v,h,t) hm v:thws ,Tf IL ('tpproxi|nnt<'ly
tlnily) "tr'l' Sllecili{'d .ttl_d I+m_ tr'dling <'<lg<,sat{' p{,rmissil>l< lli_zh sul)sonie
or lr:tnsonic bhtdit_g, 'ts would 1,' imlival{'d by the, discussi+m in lh<, next
Imr:_._'rnph, uses lhis lyl)l, +_fsurf')+'+' in ord<,r 1+t pr_'w,nt flow aced_,rzttion
on the lail +7t' th+, blade, :tn<l ku+,p lh_, :_ssouinl_,d l+_ss+,s low. t_rin(.ipa]
l)rold_'ms x_ilh :t sl]'ai_,ht surf,:te(' Dl'(+lhltl )hi" low ])s v'lhlus lTr<'cht<t(' Iow-
solidily d<,sig_ns and lh(' hmg lrailing +'dff,('e:tn 1)(,u_m_u st fuel ttralty flimsy..\l+)st <'_mv+'nli+Tnal g:_.s-tur'llin(' bhuling ulilizes sore+' nmounl of curva-
l)lr(' }_'t w('<'n Ill<' lh)'{,al "tnd l l'nilin_-+,(l_, r+,Ki<m. This l)+'rl_fits s,ml_' dif-
t'usi,m and :uldili+ma] hmdin_ _t) lh+' tail +_tlh+' bl'uh +,and it adds e,msi&,r-
:d)lv t<_ the slru<'ltmll int_,_rily <_flh{' Ida&, 1)3" in) r,_dtt<'in_ :t w_'d_:' angh,
:tl lh_' i,xil. If wmw,rdion:d]3: hm&'d bhtding is us_,d, lh{, +'ff<'(!l of this
('urw'd surf:u., _,t) l,,ss is n.+ _r-,al..ks il,ti+':tl,,d 173 figure 4 13 (which is
fr',,n_ r_q'. £), if lh<' i'xil-flow ,_[auh nttmbur is l+,ss than O.,R, th_+ ettrvalurl,
_,ff+.¢'tis snml]..\t high+,r ,xil 3I:t('1) ntmd>_,rs (ff,r+,at(,r than 0.£), th+' _,ft'uet(111 ]+lSS I":111 })('1"111111' N('V('F('. Wh(q'(,f(,l'(,, d{,sigI, ('11I'V:11 l|Fl':"q S}IOUI(1 })(' ]+_'_\'('F ill
th(' hig;h¢'r _\l:l('h mnnt_(,r r'_'g:icms."l'h_, 13p_, <d' cttrvnttm, s¢,l_.(.l_,d for ih_'
sueli+msurf:w_,l.dw_,_,n(h_'throat_nd Irniling_'_Ig<'}ms nr)+'ff_'+'t.n the
stteli,m-st|rf:w¢, vdo+'ity dislribution. In g+,n<'r'tl, th+' v_'I_mity distribution
is ilnpt'_ved if lh+' cttrvalur<, th'er+,as+,s l'r+m_ throat 1o trnilir_ ('dg_' inst+'adof rvtnainit)ff, e¢>nslal!l.
21
'I'I:I{I{INI,_
4 --
r_
2
ANI) APPIA('ATI_)N
Machnumber
1.0
_ .9
1 I I _'61.2 .4 .6 .8
Ratio of bladespacing to surface radius of curvature
Fl(;Ultl: | 1;{. V:tri:tli,m of t)r(dile loss with _la('h ntilllltvr :tnll slll'f:tce i.tlrv:ittlre
Imtvceen lhl'l)at :tn(] exit (frlml re]'. _).
Inlet
The h,ading-e(lg(' ge.nwtry of a turlfin(, bit,(l(' r.w is usually less critical
than th,, {,xil-rr_i(m _eometry. At tit{, 14:uh' inh't, n rd'ttivdy large
l(,ading-odgr r:t(litls ('Jill tlStl't]ly })(' I|s('(l, b('caus(' t}l¢' .\[:tr'h lltlllli)('r is g.n-
erally h)w at the inh,t and then iner('asos throuAh thr 14a(h' rmv. Tlw lead-
ing e(lg(' I)(,romos "t serious (umc(,rn fl)r h_w-r(,a('ti(m blading and high _\lach
number blading, in the case of h.w-r<,,'ti,,n Idacling, vxcessiv(,ly high
vt,lo('itic's in the inh,l region ran load to high v'fluos -f su(qion-surface
diffusi.u tllld _t tendency toward i:|crt-ls(,d ]()ssos. \Vith high inlet, Nlach
llllllI])('l"S, (';ll'(' III/ISt })0 t:Ik('n (}mf the 'tr('a ('ontr:t('ti(m is n()t so sever(" as
to ('lmke lh(' })]tld(' tit t]l(' inh't, l.:qualions (4 26) lind (4 27), which were
used t', w the bl./(h, exit, can also be usod t,) dotormin, a blado-inh't op(,ning
and "wilhin-lho-blado" flmv anggh' and .\lath mmflwr to cheek for blade-
inlet ('h.king.
Allh,m,,z.h circul;n" l(,ading; edgos aro usually specilh'd, this is arbitrary
and c.uld limit th(' fre(,d()ln ()f vvh_rity-distritmti(.l s('le('tion in the
h':tding-('dge r(,gion. Tho larg:e ('au'v:ttur('s ass(wi:ttod with ('irrular h"tding
rd_os (':m result in und('siral)le v('hwity l)('al.:s .n b(,th tim su('ticm- "rod
l)r(.ssur(,-sm.f:t..o i),wli(ms ()f the h,'l(ling (,(t_('. ()th(,r _(,(m.,tri('s, su('h as
dlil)s('s, whi.h p,'rmit variat ions in curval ur. "rotund tt.' h,ading: .'(t_z(', can
t)e us('(t to minimize (w (,limiwlt(' th(, w,l,,('ily p(,nks.
Blade-Surface Profile
()rico th(' leading- and trailing-cdg(' go,,m('tri('s h:tvr t)(,(,n s(,h'('t('d, the
task re n fiuing is to join dwm with a 1)r(,lih' that yi('lds thl" r(,quir('d flow
22
liI.'_,I)E I)E_I(;N
turning and a satisfactory velocity distribution around the t)]ade. The
design procedure must describe the flow conditions through the blade rows
to an accuracy sufficient to impose design controls (e.g., diffusion limits).
Two of the major flow considerations are illustrated in figure 4-14.
V(,locity gradients occur across the channel from the suction to thc
pressure surface as a result ()f the static-pressure difference required to
turn the flow. Radial variations in streamline position and, ther(,f()re,velocity occur as a result of radial-equilibrium considerations. Since both
of th(,se factors influence the blade-surface velocity distribution, the designprocedures used should be at least of a (tuasi-thre(,-dimensional nature.
The channel flow analysis theory that, serves as the basis for these design
procedures and the comput(,r programs available t() pcrfl)rm the com-
putations are discussed in the next chapter.
Pressure T Suction
surface 7 _surface
'i* /
\
Flow"_
(a)
surface surface_...
Cross-channel distance
.:_._::_:_:_:;:;- I ::;:::::::::::::::: ::::::.
_tator _/i] ', I/i_}/Rotorfiiil
Flow ' ":_''_'''_'': ::';:::::::::"t " '_'':i'_!'_:!:i_'_'_ .y.._.. ..::_.:
t ,.. "'" :."
Tip
Hub
(b)
I "-Tip
,-Hub!
Velocity
(a) Cross-chamlcl variation.
(b) Radial wtriation.
FIGURE 4-14. -Turbine bl'zde-r.w velocity variations.
23
TURBINE DESIGN AND APPLICATION
REFERENCES
1. ZWEIFEL, O.: The Spacing of Turbo-Machine Bla(ting, Especi.dly with Large
Angular l)efleeti.n. Brown t3overi Rev., w_l. 32, no. 12, Dec. 11.}45, pp. 436-444.
2. LII"BLEIN, SEYMOUB; SCHWENK, FRANCIS C.; AND ]:_]_ODERICK, ]_OBERT L. : l)if_ll-
si(m Faet(_r f_)r Estimating Losses "m([ Limiting 131ade Loadings ill Axial-Flow-
Conq)ressor 131ade Elements. NACA RM E531)O1, 11.)53.
_. STF;WART, WARNI';It L.; GLASSMAN, ARTHUIt J.; AND VANCO, _ICIIAE], R.: Ux-
aminati_m (>f Axi.d-Fh)w Turbine P, la(te-L()a<ting Charaeteristi<'s Using l)iffusi<m
Parameters. P:q)er 67-WA/GT-S, ASME, Nov. 1967.
4. STEWART, WAI_,NEI¢. L.; WHrrNt,:v, WARREN J.; AND MISEIL JAMES W.: ]Jse of
Effective M.metltum Thieklmss in l)eseril)ing Turl)ilm R.tor-l?,l'tde Losses.
NACA RM E5111_,21.), 1956.
5. WON(], ]{Oll|';RT Y.; AND S'I'F'WAItT, WAltNF'R L.: Correiati.n (ff Turbine-Blade-
Eleme,,t L_)sses l_,ase(l (m Wake M_)mentum Thickness with 1 )iffusi_m Parameter
f.r a Series of Sul)so,fie Turhil_e Blades in Two-l)imellsi()nal Cas('a(le and for
F(mr "['ral_s,)ni(' Turbine llot(_rs. NACA RM E551_;O_,, 1955.
6. Mls_.:_t, JAMn._ W.; S'rnwAwr, WAICNI':,¢ I,.; ,._Nt) W}IITNEY, WAILRF_'N J.: Analysis
of "l'urhoma(,hine Viscous Losses Affeete(t I)y ChaLlges in I?,lade Geometry.
NACA RM EStIF21, 1951i.
7. HELL_'I_, JA('K A.; W]tI']'NI';Y, l{osl,; L.; AND Cxvl(:('l_I, Rwn _l_u 11. : Exl)eriment'd
lnvestig;itiml .f a Conservatively l)esig_xell Turbine :it Four Rotor-131ade
S.lidities. NACA RM E52C17, 1952.
_. AINLI.:Y, 1). (;.; AND _IATHIFSON, (]1. C. I{.: All Examination of the Flow and
Pressure Lo_ses in ]_,lade Rows of Axial-FhJw Turl)ines. Ret). R&M 2891,
Aer(mauti_,:d F, ese:lr(.h Council, Gt. l?,rit'dll, 1955.
0. I_JET'rNI.:It, JAMI.:_ L. ; AND N(L_EK, STAN1,FY _(.: Summary of Tests on Two Highly
Loaded Turhil_e P, la(te Cotx('epts i. Three-I)imension:d C'_s("t(te Sector. Paper
1i9 WA/GT-5, AS.ME, N()v. 19(i(.L
10. Lcl,:oF:_ts, It. G.; AND ROr:LKF:, R. J.: S_)me Experime_tal Results of Two Con-
eepts l)esigned to In('re:lse "l'urlfi_xe Bl:lde Loading. J. E_g. Power, vol. 92, no.
2, Apr. 11.)70, lip. l(.)S-2()(i.
11. STABE, ROY (;.: I)esigu "lnl[ Tw()-l)imel_sion:d Cascade Test (ff Turbine Stator
Blade with t_.atio (,f Axi:d Ch.rd to Sl)a('itxg of 0.5. NASA TM X-1991, 1970.
12, Nos_,:K, STANLF:Y M.; AN/) KI.INI,:, JOHN F.: Tw[)-l)ime.si_._al Casea(te Investiga-
ti(m of a "l'urhi_m T:lnllem Blade 1)esig_l. NASA TM X-l_3(i, 191)9.
13. Nosr:K, S'rANI.r:Y M.; AND KIANF:, JOHN F.: Tw(_-l)ime_si_m'd C.tse.tde Test of a
,Jet-Flap Turl)ine l(()t[)r l',hlde. NASA TM X 21S3, 1971.
14. STA.F:, RoY (;. : l )esigll :lll(t Tw.-1 )imensi.nal Cascade Test of a Jet-Flap Turbine
Stator Bl:_([e with R:_ti_ of Axial Chord t[) Spacing of 0.5. NASA TM X-2426,
1971.
15. ]_ETTNER, JAME_ L.: l)e_ign :roll Exl)erime_ltnl Results (_f a ltighly Loaded, Low
S_)li(lity "l':_n(tem Rotor. NASA Cl¢-lS0?,, 1971.
16. ]_E'rTNEI¢ 3AME._ L.: 1)esign ami Experimental Results of a tlighly Loaded, I/ow
Solidity, Jet Fl:q_ Rot_w. NASA CR-I(,)(iS, 11.)72.
17. I)RUST, tIH_M_N W., Jl¢.; AND HrLON, I{ONALD M.: Effect of Trailing-Edge
(le[m_etry and "l'hi(,kl_ess on the Pert'orm:lnee of Cert:lin Turl)ine Stator Bl'td-
inK. NASA TN I)-6637, 11.)72.
24
BLADE DESIGN
SYMBOLS
A flow area, m2; ft 2
c chord, m; ft
D diffusion parameter
F force, N; lb
g conversion constant, 1; 32.17 (Ibm) (ft)/(lbf) (sec 2)
K ratio of inlet to exit tangential components of velocity ( V, .l/V,,.2)o throat opening, m; ft
p absolute pressure, N/m2; lb/ft _R reaction
s blade spacing, m; ft
t trailing-edge thickness, m; ft
V absolute velocity, m/see; ft/sec
x axial distance, m; ft
a fluid absolute angle from axial direction, deg
a, blade stagger angle from axial direction, deg
ratio of specific heat at constant pressure to specific heat atconstant volume
p density, kg/m3; lb/Wsolidity
_b loadiBg coefficient defined by equation (4-6)
¢_, loading coefficient defined by equation (4-5)
Subscripts:
cr critical
inc incompressiblemax maximum value
rain minimum value
opt optimum
p pressure surfaces suction surface
ss supersonic
u tangential component
x axial component1 blade row inlet
2 blade row exit
2a within trailing edge of blade row
Superscript:
absolute total state
25
i
CHAPTER5
ChannelFlowAnalysis
By TheodoreKatsanis
The design of a proper blade profile, as indicated in the last section of
chapter 4, requires calculation of the blade-row flow field in order to
determine the velocities on the blade surfaces. This chapter presents theanalysis theory for several methods used for this calculation and also
discusses associated computer programs that were developed at NASALewis Research Center.
The actual velocity distribution throughout a blade-row flow field
cannot be calculated at this time because of the extreme complexity of
nonsteady, viscous, three-dimensional flow through geometl'ically complex
passages. To calculate a theoretical velocity distribution, therefore, certainsimplifying assumptions must be made. The three-dimensional flow is
simplified to flow on or through various two-dimensional surfaces. Such
surfaces are illustrated in figure 5-1 for the case of a radial-inflow turbine.Similar surfaces are used for an axial-flo(v turbine. A flow solution on the
mean hub-to-shroud stream surface (commonly called the meridional
surface), shown in figure 5-1(a), does not yield blade-surface velocities
directly, but provides information required for the blade-to-blade surface
(fig. 5-1 (b)) and orthogonal surface (fig. 5-1 (c)) solutions, which yieldthe desired blade-surface velocities.
There are two parts to a method of analysis to obtain a velocity dis-tribution over one of these surfaces. The first part is the mathematical
formulation of the problem, and the second part is the numerical solution
of the mathematical problem. For the mathematical formulation of the
problem, we will discuss stream- and potential-function methods and
velocity-gradient (stream-filament) methods. The stream- and potential-
27
t
TURBINE DESIGN AND APPLICATION
Hub-to-shroudstreamsurface-_
?
(a)
Bla(le-to-bla(le
surface7
,- 0 rthogonalsurface
't(¢)
(a) Hub-to-shroud stream surface. (b) Blade-to-blade surface.(c) Orthogonal surface across flow passage.
FmURE 5-1.--Surfaces used for velocity-distribution calculations.
28
CHANNEL FLOW ANALYSIS
function methods will be described relative to the blade-to-blade surface
solution. A similar type of analysis can be made for the meridional surface.
The velocity-gradient equation to be presented is general and can be usedfor solutions on any of the surfaces.
The following assumptions are made in deriving the various methodsof analysis discussed herein:
(1) The flow is steady relative to the blade. This means that the
surface velocity at any given point on the blade does not vary with time.
Thus, if the blade is rotating, the flow would not be steady relative to afixed coordinate system.
(2) The fluid obeys the ideal-gas law
p=pRT (5-1)where
p absolute pressure, N/mS; lb/ft 2p density, kg/m_; lb/ft 3
R gas constant, J/(kg) (K) ; (ft) (lbf)/(lbm) (°R)T absolute temperature, K; °R
or is incompressible (p = constant).
(3) The fluid is nonviscous. A nonviscous fluid has no boundary layer.The blade-surface velocity is calculated, therefore, as if the free streamextends to the blade surface.
(4) The fluid has a constant heat capacity.(5) The flow is isentropic.
(6) The total temperature and total pressure are uniform across theinlet.
(7) For the stream- and potential-function analyses, the additional
assumption is made that the flow is absolutely irrotational. Therefore,
curl V = VX V = 0 (5-2)
where V is the absolute velocity vector. Intuitively, this means that
particles do not change their absolute orientation with time, although
their shape may change. For example, figure 5-2 shows a hypothetical
particle at times t and t+At. In the absolute frame of reference, the
particle changes its location and shape at a later instant of time, but thenet rotation is zero. Of course, in a frame of reference relative to the
blade, the particle has rotated, because the frame of reference has rotated.
29
TURBINE DESIGN AND APPLICATION
Direction of rotation
Time = t
_Time= t + At I I
(Absolute frame of reference)
Time = t + At
(Relative frame of reference)
FIGURE 5--2.--Absolutely irrotational flow.
Some numerical techniques for solving the mathematical equations will
also be discussed. However, it must be emphasized that there are many
techniques for solving these equations, and we will discuss only a few. Anexcellent theoretical discussion of flow in two-dimensional cascades is
given in Chapter IV of reference 1.
STREAM- AND POTENTIAL-FUNCTION ANALYSES
Stream-Function Method
The stream function can be defined several ways, but perhaps the
simplest is in terms of streamlines. Suppose we consider two blades of acascade as shown in figure 5-3. It is assumed that there is two-dimensional
3O
CHANNEL FLOW ANALYSIS
Mass
flow
fractbn
.8t
.et
.41
o
FIGURE5-3.--Streamlines for a stator cascade.
axial flow here, so that the radius r from the centerline is constant and
there is no variation of the flow in the radial direction. There may berotation about the centerline.
Shown in figure 5-3 are a number of streamlines. The mass flow between
the blades is w. The number by each streamline indicates the fraction of w
passing between the upper surface of the lower blade and the given stream-
line. Thus, the upper surface (which is a streamline) has the value 0, and
the lower surface of the upper blade has the value of 1, while the remainingstreamlines have values between 0 and 1. Note that a value can be asso-ciated with any point in the passage. This value is called the stream-
function value and can be used to define thc stream function.
It will be recalled that mass flow can be calculated for a one-dimensional(or uniform) flow by
w=pVA (5-3)
31
TURBINE DESIGN AND APPLICATION
where
w rate of mass flow, kg/sec; lb/sec
V fluid absolute velocity, m/sec; ft/secA flow area normal to the direction of the velocity V, m2; ft _
This can be extended to a varying flow by using an integral expression:
w=f, pvdA (5-4)
Since this stream-function analysis applies to both stationary and rotating
cascades (blade rows), the fluid velocity will be expressed in terms of
relative velocity W, which for a stationary blade row reduces to absolute
velocity V. We will assume that our cascade has a uniform height b. Then,
the mass flow wl._ between any two points QI and Q2 in the passage (see
fig. 5-4) can be calculated by
pW,,bdq (5-5)
-/-JQ1
FIGURE 5-4.--Arbitrary curve joining two points in flow passage.
32
CHANNEL FLOW ANALYSIS
where W, is the relative velocity component in the direction of the right-hand normal of the line going from Q_ to Q_. This sign convention means
that wi.2 will be negative if Q2 is below a streamline passing through QI.The integral is a line integral between the points Q_ and Q_ and is in-dependent of path for steady flow relative to the cascade.
With the use of equation (5-5), an analytical expression can be writtenfor the stream function u at a point (x, y) :
/Q("'Y) pW,,b dqo
u(x, y) = (5-6)W
where Q0 is any point on the upper surface of the lower blade, and theintegral is taken along any curve between Q0 and (x, y). This is indicatedin figure 5-5.
Since the integral in equation (5-6) is independent of path, it is rela-
tively easy to calculate the partial derivatives of u. For example, we will
calculate Ou/Ox at the point (x, y). Let x0<x such that the point (Xo, y)is still in the flow passage, as shown in figure 5-6. Then
fc_ pW,,b dqT /c _ pW,b dq
u(x, y) =W
(5-7)
FIGURE 5-5.--Curve joining (x,y) with a point on the upper surface of the lower blade,
33
TURBINE DESIGN AND APPLICATION
(x o, Y)/- C2, (x,y)
Wy _ W
FIGURE 5-6.--Curve joining horizontal line through (x,y) with a point on the uppersurface of the lower blade.
where C1 is an arbitrary curve between Q0 and (x0, y), and C2 is a hori-
zontal line between (x0, y) and (x, y). The integral along C1 does not
depend on x. Along C,, we have W, = -Wv and dq = dx. Hence,
0 f; pW_b dx (5-8)0U(x,y)=__xx o wOx
or
Ou_= _ pW_b (5-9)cOx w
In a similar manner, we can calculate
cOu=pW=b (5-10)
Oy w
Now we will make use of the fact that the flow is absolutely irrotational.
From the definition of the curl operator and the above assumption,
curlV=\_yy _-z/iW_z Ox/J-F_k_x _Y] =0 (5-11)
wherei, j, and k are the unit vectors in the x, y, and z directions, respec-
34
CHANNEL FLOW ANALYSIS
tively, and Vx, V_, and V, are the absolute velocity components (in
m/see or ft/sec) in the x, y, and z directions, respectively. Since we are
considering two-dimensional flow only,
V,=O (5-12)and
0 V, 0V_--- =0 (5-13Oz Oz
Hence, equation (5-11) requires only that
Since
and
OV_ OV:-- = (5-14)Ox Oy
v_=wx (5-15)
Vv = Wu+o_r (5-16
where ¢o is the angular speed (in rad/see) and the radius r is constant,equation (5-14) can be expressed in terms of relative velocities as
OW u OW,
Ox Oy(5-17)
Actually, the flow is irrotational with respect to the moving coordinates
in this particular case. Now, from equations (5-10) and (5-9),
w Ou
Wx = pb Oy (5-18)
w Ou
w,=-o-gox (5-19)
Substituting equations (5-18) and (5-19) into equation (5-17) yields
0 (10u\ 0 (lOu_ 0ox ,,; (5-20)since w and b are both constant.
For incompressible flow, p is constant, and
02u O_u
V2u = Ox_+Oy_ = 0 (5-21)
which is Laplace's equation. Any function satisfying Laplace's equation is
called a harmonic function. There is a great deal of theory concerning
35
TURBINE DESIGN AND APPLICATION
harmonic functions that is related to the theory of analytic functions of
complex variables.The important thing to know here is that there are a tremendous
number of functions that satisfy equation (5-21), and we must find a
solution that satisfies certain boundary conditions. The solution to either
Laplace's equation (5-21) or equation (5-20) will be determined by
specifying two things: (1) a finite region, and (2) a boundary condition
along the entire boundary of the region.
The first thing that must be specified is the solution region. A typical
two-dimensional cascade is shown in figure 5-7. Since the flow is the samein every passage, we can consider a finite solution region as shown in
figure 5-8. It is assumed that AH is sufficiently far upstream so that the
flow is uniform along this part of the boundary and that the flow angle
/_, is known. Similarly, it is assumed that the flow is uniform along
DE, and that the flow angle/_o_t is known. From the way the stream
function was defined, we can specify boundary conditions on the entire
boundary ABCDEFGHA. Along BC, u=0; and along FG, u-- 1. Along
AB, HG, CD, and FE, a periodic condition exists; that is, the value of u
along HG and FE is exactly 1 greater than it is along AB and CD. Along
AH and DE, Ou/O_ is known, where _ is the distance in the direction ofthe outer normal.
Fmu_ 5-7.--Two-dimensional infinite cascade.
x
36
CHANNEL FLOW ANALYSIS
H
Unil°rml l
n
AY
FIGURE 5--8.--Finite solution region.
Jniform
flow
_o-out
W
Consider the differential of u in the direction of the velocity W:
du = au dx-k_y dy = 0 (5-22)ax
The differential is 0 because the stream function is constant along astreamline, and the velocity vector must be tangent to a streamline.Along AH,
OU OU
07 Ox
and substitution from equation (5-22) yields
(5-23)
However,
Ou Ou dy
07 - Oy dx (5-24)
dyd-_ = tan B (5-25)
Further, Ou/Oy is constant along AH, since it is assumed that the flow isuniform there. Therefore,
Ou [u(H)-u(A)] 1
Oy s s (5-26)
where s is the blade spacing in the y direction. Substituting equations(5-25) and (5-26) in equation (5-24) gives along AH
,. = (5-27)8
37
TURBINE DESIGN AND APPLICATION
Similarly, along DE, one can calculate
0___t_ tan _out (5-28)O_l lout S
We now have a boundary condition along the entire boundary of the region
shown in figure 5-8. These boundary conditions will always determine a
unique solution to Laplace's equation (5-21). For compressible flow (eq.
(5-20)), a unique solution is always determined if the flow is strictly
subsonic throughout the region.
There are numerous techniques for solving equation (5-20) or (5-21).
After the stream function is obtained, blade-surface velocities and velocities
throughout the passage can be obtained by differentiation of the stream
function. This is what is known as the direct problem. A method of solving
this problem will be discussed later. The indirect, or inverse, problem is to
specify a desired velocity distribution on the blade surface and from thisdetermine a blade shape that will give this velocity distribution. This willnot be discussed here.
Potential-Function Method
For two-dimensional irrotational fow, a potential function can be
defined. If lines of equal potential are drawn, they will be orthogonal to
streamlines. The potential function will not be defined in the same detail
as the stream function, but the main properties and relations will be given.
If the potential function • exists (i.e., the flow is irrotational), then it canbe defined so that
and
04--=V, (5-29)0x
04--= V_ (5-30)Oy
We will refer to absolute velocities here, since we must have flow irrota-
tional relative to the coordinate system used. This, coupled with the
assumption of absolute irrotational flow, implies that the coordinate
system does not rotate. This does not exclude use of the potential function
for pure axial flow, since the rotation has no effect if there is no change in
radius; that is, the flow is actually irrotational with respect to the blades,as we saw in the discussion of the stream function.
From the continuity relationship for steady flow,
o(pv2 o(pv_)-_ 0 (5-31)
Ox Oy
38
CHANNEL FLO% _ ANALYSIS
Substituting equations (5-29) and (5-30) in equation (5-31) yields
0 (5-32)If the flow is incompressible, p is constant, and
02_ 02_
v2®= =0 (5-33)
So, the potential function satisfies Laplace's equation. Thus, for incom-
pressible, irrotational flow, both the stream function and the potential
function satisfy the same differential equation (Laplace's equation). Thedifference lies in the boundary conditions.
We can consider the same solution region shown in figure 5-8. We can
specify boundary conditions over the entire boundary as follows: AlongBC and FG,
- Vn =0 (5-34)07
where Vn is the velocity normal to the blade surface. Along AH,
in
and along DE,
out
The inlet and outlet axial velocities are given by the equations
W
(Vx),, p,_bs (5-37)
and
W
(V_)o_, po_tbs (5--38)
Along AB, GH, CD, and EF, a periodic condition exists. Since the flow isuniform along AH,
Substituting
(0yy_) [_(H)- q,(A)-] (V_) _. (5-39)8
V, = V, tan B (5-40)
39
TURBINE DESIGN AND APPLICATION
into equation (5-39) yields
¢(H) = _(A) Ts( V_),. tan _,n (5-41)
Because of the periodicity, • is exactly s(V_)_. tan f_ greater along HG
than along AB. Similarly, at the outlet,
4_(E) = ¢(D) _-s(V_)o,,_ tan f_o_, (5-42)
Equation (5-42) gives the difference in • along the lines FE and CD.
This completes the boundary conditions for equation (5-32) or (5-33).
The boundary conditions, however, do not determine a unique solution,but only a solution within an arbitrary additive constant. If the value of
¢ is specified at one point, these boundary conditions will determine a
unique solution to equation (5-33) for incompressible flow, or to equation
(5-32), for strictly subsonic compressible flow throughout the region.As for the stream function, there are numerous methods for solving
equation (5-32) or (5-33) subject to the preceding or equivalent bound-
ary conditions. A method for solving the inverse problem of specifying
the velocity distribution to determine the blade shape is described inreferences 2 and 3.
Choice of Stream- or Potential-Function Method
If the flow is steady, irrotational, and incompressible, there is little to
choose between the stream function and the potential function. In this
case, the choice is made on the basis of ease of solution for the boundary
conditions (the differential equation is the same: Laplace's equation).
However, if any of the three assumptions (steady, irrotational, or incom-
pressible flow) is not applicable, then we may be restricted as to thechoice of stream function or potential function.
The existence of the stream function is proven from the continuity
equation. For the stream function to be defined, the mass flow crossing
a line between two points must be independent of path. This requires
that the flow be either incompressible or steady. Some additional assump-
tion is necessary for the flow to be unique. We used the assumption that
the flow was absolutely irrotational, which turned out to be irrotational
relative to the blade for the axial-flow case considered. However, other
assumptions could be made for other problems. Another restriction onthe stream function is that it can be defined only for two-dimensional flow.
This can easily be seen since the stream function is defined as a percentage
of mass flow between two points, and this is meaningless in threedimensions.
The existence of the potential function can be shown if the flow isirrotational relative to the given coordinate system. This is necessary
4O
CHANNEL FLOW ANALYSIS
because we must have equality of mixed second partial derivatives; thatis, if
0_ a2_
axay-oyox (5-43)
then
0 V -0 V. (5--44)Ox _-Oy
and the flow must be irrotational. A similar situation exists in three-
dimensional flow; that is, the potential function exists only if the flow is
irrotational with respect to the coordinate system being used. Finally,an assumption must be made to assure a unique solution. This can bedone by using the continuity equation.
Finite-Difference Solution for Stream-Function Method
As stated before, there are many ways of solving various problems
posed by stream-function or potential-function theory. We "will consider
in further detail the finite-difference solution of the direct problem for the
stream function for the simplest case of steady, incompressible, irrotationalflow. In this case, we must solve Laplacc's equation subject to the bound-
ary conditions discussed in the section on. the stream function. The method
of solution for the potential function is quite similar, but with a lowerrate of convergence for the finite difference solution.
The first step is to establish a rectangular grid of mesh points in theregion shown in figure 5-8. A typical grid is shown in figure 5-9. Then a
finite-difference approximation to Laplace's equation (eq. (5-21)) can bewritten at each mesh point where the stream function is unknown. A
typical mesh point with four neighboring mesh points is shown in figure
5-10. The point in consideration is labeled 0, and the four neighboringpoints are labeled 1 to 4, as shown. The distance between points 1 and 0 is
denoted ]11, and similarly, the other distances are h2, h3, and h4 as indicated
in figure 5-10. The value of u at points 0 to 4 are labeled u0 to u4, respec-tively. With the use of a Taylor series expansion for u in the x- and y-direc-
tions, equation (5-21) can be approximated by using only values of u at
mesh points. (Further explanation of this is given in ch. 6 of ref. 4.)When this is done, the following expression is obtained:
2Ul 2u_
h_(hx-k-h_) -{ h_(h_Th2)2Uo ] ]- 2U3 2tt4 2Uo ] 0h-_,J+[hn(h-_h4)-_ h,(ha+h,) h_,J =
(5-45)
41
TURBINE DESIGN AND APPLICATION
I J I
I [ J
J I Ili i
[ i I
, Ill ]
FIGURE 5-9.--Mesh used for a finite-difference solution.
]D
h 3
h2
h4-4
h 1
FmVaE 5-10.--Notation for adjacent mesh points and mesh spaces.
42
CHANNEL FLOW ANALYSIS
Solving equation (5-45) for u0 yields the expression
4
Uo=Zaiui (5-46)
i-i
where
h_-{-h4al - (5-47)
aohl
h3+h4a, - (5-48)
aoh2
ha+h_a3 = -- (5-49)
aoh3
hlA-h2a4 - (5-50)
aoh4
1 1 1 1
ao=(h3+h4) (_+_)+(h'+h') (h3+h4)(5-51)
Equation (5-46) holds at every interior mesh point. If one of the neigh-
boring points is on a blade surface, then the value of u at that point can beused. At other points along the boundary, equation (5-46) cannot be used,
but the boundary conditions can be used to obtain alternate equations at
these points. For example, along the upstream boundary AH in figure 5-9,
Ou/On is given by equation (5-27). If point 0 is on line AH, then, a finite
difference approximation gives
(5-52)
Similarly, if point 0 is on line DE,
Uo= u3- h3 (tan_ °_) (5-53)
For the points along AB and CD, equations can be derived by using
the periodic boundary condition. If the point 0 (fig. 5-11) is on the
boundary between A and B, the point 1 is outside the boundary. However,it is known that ul=ul.,-1, where the point 1,s is a distance s above
point 1 in the y-direction, as shown in figure 5-11. Substituting this
43
TURBINE DESIGN AND APPLICATION
H 2
I
u2 " u2,-s + 1
A 2,-s B
FIGUaE 5--ll.--Mesh point on line AB.
condition in equation (5--46) gives
4
uo=alul,t,q-- _ aiui--al (5--54)i--2
This equation holds along CD (fig. 5-8) also.
The points along HG need not be considered, since they are just 1
greater than the corresponding point along AB. The equation for the firstmesh line below HG, therefore, must be modified, since point 2 is on line
HG. In this case, ua=u2._,+l, where the point 2,-s is a distance s
below point 2 in the negative y-direction, as indicated in figure 5-12.Substituting this condition in equation (5--46) gives
Uo= alul-t- anu2._,-_ asu8 + a4u4 -J- a2 (5--55)
This equation also applies to the first mesh line below FE (fig. 5--8).One of equations (5--46) or (5-52) to (5-55) can be applied to each
mesh point for which the stream function is unknown in the region ofinterest to give the same number of linear equations as there are un-
knowns. These points where the stream function is unknown will bereferred to simply as unknown mesh points.
Suppose that there are n unknown mesh points. We then have n equa-tions in n unknowns. The points can be numbered consecutively from 1 to
n. The values of u will then be ul at the first point, u2 at the second point,
and so forth up to u. at the last point. At each point, one equation willapply. The equation at a typical point, i, could be written
44
CHANNEL FLOW ANALYSIS
G.¢
f
Ul-Ul, s- 1
A B
FIouaE 5-12.--Mesh point on first line below HG.
n
__, aljui = kl (5-56)j--1
The values of the aij are determined by one of equations (5-47) through
(5-55). All but five, at most, of the aij are zero, and the all=- 1. The
value of ki is always zero, except for the outermost unknown points
around the boundary. It can be shown that the alj matrix is always non-
singular; hence there is always a unique solution for the uj.
A numerical solution to equation (5-56) can be obtained by iterative
techniques. These techniques are particularly valuable in solving systems
of linear equations of this type; that is, where there are a large number ofunknowns, but few terms in each equation. Storage requirements are
small, and roundoff error is minimized with iterative methods. To start
the iteration, an initial estimate of u at every unknown mesh point is
required. The simplest iterative procedure is relaxation. This consists ofchanging the estimated value of u at each point in succession so as to
satisfy the equation for that point. After this is done at every point, the
procedure is repeated until there is negligible change in the values of u.
The procedure is simple and it always converges for this problem. How-
ever, the convergence rate is extremely slow, so that exci_ssive computer
time is required. The convergence can be accelerated greatly by increasing
the change in u at each iteration by a factor o:, called the overrelaxation
factor. When ¢0= 1, the procedure is straight relaxation, and when oJ> 1,it is overrelaxation. It is proven in reference 4 that overrelaxation (or
underrelaxation) is convergent if 0 < _0< 2. However, the greatest rate of
convergence occurs when 1 <w<2. In fact, there is an optimum value of _o
45
TURBINE DESIGN AND APPLICATION
between 1 and 2 which gives the most rapid convergence. This optimumoverrelaxation factor can be calculated as explained in reference 4.
To give an explicit expression for the overrelaxation procedure, we willuse a superscript on the us. That is, uV" is the m th iterate of ul. The initial
estimates are denoted u? and may be any value. For example, an initial
estimate of u?=O is satisfactory. Then, if u__ is known for all i, we cancalculate ul m+_, for i= 1, 2,..., n in succession by
}_im÷l = uim2v¢O -- _ ai_uj re+l- aijujm+kl--ui mj-i j-i+l
(5-57)
After a solution for u is obtained by overrelaxation (or any other
method), it is necessary to calculate the velocities with the use of equa-tions (5-9) and (5-10) as
Wz
pb(5-58)
and
w(O;)W_ = (5-59)
pb
The partial derivatives Ou/Ox and Ou/Oy must be estimated from the
calculated discrete values of ui. This can be readily done, either by finite
differences, or by fitting a smooth curve, such as a spline curve, through
the points. The resultant velocity is calculated from the two components
at unknown mesh points. On the blade surface, the velocity is calculatedfrom one component and the blade tangent angle.
Computer Programs for Stream-Function Analyses
As can be seen, the solution of Laplace's equation and the calculation
of velocities is a lengthy calculation procedure which is best done by com-puter. Several computer programs have been written at the NASA Lewis
Research Center for the analysis of flow through turbomachine bladingby stream-function methods. Most of these programs are for blade-to-
blade analysis (region shown in fig. 5-9). The program called TURBLE,
which is described in reference 5, can be used to analyze axial, radial, ormixed flow. In accordance with the constraints associated with the
stream-function method, the flow must be subsonic throughout the entire
solution region. The TSONIC program, described in reference 6, super-
46
CHANNEL FLOW ANALYSIS
sedes TURBLE in that it performs all the same calculations and, in
addition, extends the solution to transonic (local supersonic velocities)
flow problems. Transonic solutions are obtained by using a velocity-gradient equation of the type described in the next section to extend a
preliminary (lower mass flow rate) subsonic stream-function solution. A
program called TANDES[, which is described in reference 7, can be usedto analyze flow in tandem or slotted blade rows or blade rows with
splitters. Another program, called MAGNFY and described in reference
8, obtains a detailed solution in the leading- or trailing-edge regions of anyblade or in the slot region of tandem or slotted blades. The TANDEM
and ._IAGNFY programs are restricted to subsonic flow.
Flow in the meridional plane (mean hub-to-shroud flow surface, as
indicated by fig. 4-14(b) or fig. 5-1(a)), of any axial- or mixed-flow
turbomachine can be analyzed by a program called 5IERIDL, which is
described in references 9 and 10. Transonic solutions can be obtained by
the use of a velocity-gradient equation to extend a preliminary subsonicstream-function solution.
VELOCITY-GRADIENT ANALYSIS
As indicated previously, the stream-function and potential-functionmethods of analysis are limited to solutions that are entirely subsonic
within the computation region. By use of a velocity-gradient equation and
additional assumptions, however, the subsonic solution can be extended to
give an approximate solution in the transonic flow regime. It is also
possible to use a velocity-gradient method of analysis alone to obtain sub-
sonic, transonic, or supersonic solutions without assumptions other than
the basic ones indicated earlier. The velocity-gradient analysis is often
called a stream-filament analysis because the velocity-gradient equation
involves the streamline, or stream-flament, curvature and position.
A velocity-gradient method of analysis can only give solutions within aguided passage; that is, a passage where both ends of all streamline
orthogonals intersect a solid boundary. Therefore, the usefulness of this
method depends on the degree of flow guidance provided by the turbine
blades. For a well-guided passage (high solidity and/or small angles),such as shown in figure 4-11, most of the suction surface is within the
guided region, and the associated surface velocity distribution can be well
defined. On the other hand, for a low-solidity blade row, such as thatshown in figure 5-9, less than half of the suction surface is within the
guided region, and surface velocities can be computed only on the front
half of the suction surface. In this latter case, the stream-function analysismust be used if better definition of the suction-surface velocity distributionis required.
47
TURBINE DESIGN AND APPLICATION
Method
The idea of a velocity-gradient method can be demonstrated by con-
sidering a simple case. Suppose we have two-dimensional flow through a
narrow passage as shown in figure 5-13. We assume the height of the
passage to be b, and the width d. If the mass flow is known, the average
velocity can be calculated approximately from continuity by
w
W,,o pbd (5--60)
However, there is a variation in velocity across the width of the passage,and in turbomachinery it is this velocity difference we are interested in.
With a force-equilibrium equation, by balancing centrifugal force against
the pressure gradient as was done in chapter 3 for consideration of radial
equilibrium, it can be shown that
dW W(5-61)
dq r,
where q is the distance from the suction (convex) surface, and rc is theradius of curvature for the streamline. The sign convention for rc is
important; r_ is positive if it is concave upward, and negative if it is con-
cave downward. For the simple case shown in figure 5-13, equation (5-61)
can be integrated along a radial line by assuming the streamline radius of
curvature to be equal in magnitude to the passage radius. There results,
for integration from the inner radius to any point in the passage,
W r, (5-62)W, r
(Flow
FIGURE 5-13.--Flow through a curved passage.
48
CHANNEL FLOW ANALYSIS
where
W,
rs
r
relative velocity on inner, or suction, surface, m/sec; ft/sec
radius of inner, or suction, surface, m; ft
radius of passage, m; ft
The mass flow through the passage is expressed as
w = pWb dr (5-63)
and substitution of equation (5-62) into (5-63) and integration, with
constant density assumed, yields
W
W, = (5-64)
pbr, In (1+ d )
In a similar manner, the outer, or pressure, surface velocity can be com-
puted as
W
Wp = (5-65)
pbr°(l_-d) ln(1-_.)
Thus, an estimate of the blade-surface velocities can be obtained simply
by using equation (5-62), which is a velocity-gradient equation. We arenot necessarily restricted to two-dimensional flow. If there were some
variation of velocity in the height of the passage, a velocity gradient couldbe calculated in that direction also.
We will now consider a very general velocity-gradient equation. Since
we are interested in turbomachinery, we will use a rotating cylindrical
coordinate system with radius r, angle O, and axis x, as shown in figure5-14. Also indicated are the velocity components, Wr, Wx, and W0. The
meridional component W, is the resultant of W, and Wx. The meridional
plane is a plane containing the x axis. Also shown in figure 5-14 are a, the
angle between W_ and the x axis, and f_, the angle between W and the
meridional plane. The following relations hold for the components:
W0 = W sin t_ (5-66)
W., = W cos _ (5--67)
W. = W_. sin a (5--68)
W. = W.. cos a (5-69)
49
TURBINE DESIGN AND APPLICATION
X
W2 . W2+ Wx2m r
FIGURE 5-14.--Cylindrical coordinate system and velocity components.
In addition to the r-, O-, and x-coordinate, it is convenient to use an
m-coordinate. The m-coordinate is the distance along a meridional stream-
line, as shown in figure 5-15. The m-distance is less than the true stream-
line distance if the angle _0. The meridional streamline is the projection
of a streamline in the meridional plane; that is, the 0-coordinate is
neglected. The curvature of the meridional streamline is 1�re, where rc is
the radius of curvature of the meridional streamline. The sign of rc is
positive if the streamline is concave upward.
We want the velocity gradient along an arbitrary curve. Let q be the
distance along this curve. For the case of constant total temperature and
constant angular momentum (rV,) at the inlet,
dW dr dx dO
dq- a Yq+c eq-- (5-7O)
where
a= wc°sac°s_/3 Wsin_/3_-sinacos/3_m--2_sin/3 (5-71)re r
5O
CHANNEL FLOW ANALYSIS
Meridional /streamline7 _"
//
r
I _Axis
_X
FIGURE5-15.--The m-coordinate.
dW_W sin a cos _ f_ _-cos a cos 3 (5-72)b_ --
re dm
/dWe'2_ sin a)c = W sin a sin 3 cos 3-_r cos 3 _-_m" (5-73)
These equations are derived as equations (B13) and (B14) of reference 11.
In using any velocity-gradient equation, it is necessary to solve a differ-
ential equation involving streamline-geometry parameters, such as cur-
vature, a, and f_. These are not known precisely in advance. However,for a guided channel, these parameters can be estimated reasonably well.
A great number of special cases can be obtained from equations (5-70)
to (5-73). For example, suppose we have an annular passage with no
blades, as shown in figure 5-16, and no velocity component in the tangen-
tial (0) direction (into page). We can calculate dW/dn, where n is thedistance normal to the streamline. Let q=n in equation (5--70). Since
We=0, then dO/dn=O and 3=0. Further, from figure 5-16, it can be
Merldlo.nal \dn_dr_ a_a J
_lnner wall
FIGURE5-16.--Annular passage with no blades.
51
TURBINE DESIGN AND APPLICATION
seen that dr/dn = cos a and dx/dn = -sin a. Then, from equations (5-70)to (5-73),
dW W
dn re (5-74)
Thus, for this case, equation (5-70) reduces to the simple form of equa-tion (5-61).
Computer Programs
Several computer programs for the analysis of flow through turbo-machine blading by velocity-gradient methods have been written at the
NASA Lewis Research Center. One that was used for many years is theCTTD program, which is described in reference 12 and is limited to axial-
flow turbines. This program has now been superseded by the more generaland easier to use CHANEL program, which is described in reference 13.
The CHANEL program can be used to analyze axial-, radial-, or mixed-flow turbines or compressors. Velocity-gradient equations are used to
determine velocity variations both from hub to tip along meridional-
streamline orthogonals and from blade to blade along hub-, mean-, and
tip-streamline orthogonals. This results in a flow solution for an orthogonalsurface, as illustrated in figure 5-17, which satisfies a specified mass flow
rate. Computations are made for a number of these surfaces along theblade passage. This program can also be used to compute the maximum
(choking) mass flow rate for the channel. The program gives good results
for medium- to high-solidity blading. As indicated previously, moredefinition than can be provided by this program may be needed for low-
solidity blading, because solutions can only be obtained for fully guidedsections of the passage.
Velocity-gradient methods have also been used to obtain meridional-
plane and blade-to-blade plane solutions. The basic method for a meridi-
onal-plane analysis for mixed-flow centrifugal impellers is presented in
reference 14, which uses the velocity-gradient equation along streamline
orthogonals. Since the orthogonal lengths are not known in advance, itwas more convenient to base a computer program on the use of the
velocity-gradient equation along fixed straight lines, which were called
quasi-orthogonals. Such programs for meridional-plane analysis are
presented in reference 11 for a radial-inflow turbine impeller and inreference 15 for backward-swept or radial impellers and vaned diffusers of
centrifugal compressors. A program for a blade-to-blade plane analysisthat uses quasi-orthogonals for a radial-inflow turbine impeller is describedin reference 16.
A further use of the velocity-gradient equation, as mentioned pre-viously in this chapter, is to extend a subsonic stream-function solution to
52
('|IANNEL FL(IW ANALYSIS
obtain local supersonic velocities. The subsonic solution is used to obtain
the flow angles and streamline curvatures required for the velocity-gradient equation. Programs for transonic-flow solutions based on this
method are presented in references 9 and 10 for a mcridional solution andin reference 6 for a blade-to-blade solution.
Orlhogonal / `/
su rface --_/
t
Tip_-to-blade ,-Suction
i surface
Mean
Parallel to axis
of rotation-_
Hub
FmVtCE 5-17.--Turhine blades with three-dimensional orthogonal surface across
flow passage.
53
TURBINE DESIGN AND APPLICATION
REFERENCES
l. JOHNSEN, IRVING A.; AND BULLOCK, ROBERT O., EDS.: Aerodynamic Design of
Axial-Flow Compressors. NASA SP-36, 1965.
2. COSTELLO, GEORGE R.: Method of Designing Cascade Blades with Prescribed
Velocity 1)istributions ill Compressible Potential Flows. NACA Rep. 978,
1950.
3. COSTELLO, GEORGE R.; CUMMINGS, ROBIJRT I,.; AND SINNETTE, JOHN T., JR.:
Detailed Computational Procedure for l)esign of Cascade Blades with Pre-
scribed Velocity l)istributions in Compressible Potential Flows. NACA Rep.
1060, 1952.
4. VARGA, RICHARD S. : Matrix Iterative Analysis. Prentice-Hall, Inc., 1962.
5. KATSANIS. THEODORE; AND MCNALIA', WILt.IAM l).: Revised FORTRAN Pro-
grain for Calculating Velocities and Streamlines on a Blade-to-Blade Stream
Surface of a Turbomachine. NASA TM X-1764, 1969.
6. KATSANIS, TtlEOI)Ot_.E: FORTRAN Program for Calculating Transonic Velocities
on a Blade-to-Blade Stream Surface of a Turbomachine. NASA TN D-5427,
1969.
7. KATSANIS, THEODORE; AND McNALLY, WII,I,IAM 1).: FORTRAN Program for
Calculating Velocities and Streamlines on a Blade-to-Blade Stream Surface of
a Tandem Blade Turt)omachine. NASA TN 1)-5044, 1969.
8. KATSANIS, TREODORE; AND _{cNALLY, WILLIAM 1).: FORTRAN Program for
Cah_ulating Velocities in a Magnified Region on a Blade-to-Blade Stream Sur-
face of a Turbomachine. NASA TN 1)-5091, 1969.
9. KATSANIS, THnODORE; AND McNAI,I,V, WILLIAM l).: FORTRAN Program for
Calculating Velocities and Streamlines on the Hub-Shroud Mid-Channel Flow
Surface of an Axial- or Mixed-Flow Turbomachine. I--User's Manual. NASA
TN I)-7343, 1973.
10. KATSANIS, THEODORE; AND McNALLY, WILLIAM l).: FORTRAN Program for
Calculating Veh)cities and Streamlines ()it the Hub-Shroud Mid-Channel Flow
Surface of an Axial- or Mixed-Flow Turbolnachine. II--Progralnmer's Manual.
NASA TN 1)-7344, 1974.
11. KATSANIS, THEODORE: Use of Arbitrary Quasi-Orthogonals for Calculating Flow
I)istribution in the Meridional Plane of a Turbomachine. NASA TN D-2546,
1964.
12. KATSANIS, THEODORE; AND DELLNER, Lois T.: A Quasi-Three-Dimensional
Method for Calculating Blade-Surface Velocities for an Axial Flow Turbine
Blade. NASA TM X-1394, 1967.
13. KATSANIS, THEODORE: FORTRAN Program for Quasi-Three-l)imensional Cal-
culatitm of Surface Velocities and Choking Flow for Turbomachine Blade Rows.
NASA TN 1)-6177, 1971.
14. HAMRICK, JOSEPH T.; GINSBURG, AMBROSE; AND OSBORNE, WALTER M. : Method
of Analysis for Compressible Flow Through Mixed-Flow Centrifugal Impellers
of Arbitrary Design. NACA Rep. 1082, 1952.
15. VANCO, MICHAEL R.: FORTRAN Program for CMculating Velocities in the
Meridional Plane of a Turbomachine. I--Centrifugal Compressor. NASA TN
D-6701, 1972.
16. KATSANIS, THEODORE: Use of Arbitrary Quasi-Orthogonals for Calculating Flow
l)istrihution on a Blade-to-Blade Surface in a Turbomachine. NASA TN
D-2809, 1965.
54
CHANNEL FLOW ANALYSIS
A
ai
bd
h
ki
m
n
P
qR
r
8
T
t
U
V
W
W
ot
P
oJ
SYMBOLS
flow area, m2; ft 2
coefficients for equation (5-46)
cascade height, m; ft
passage width, m; ft
distance between mesh points, m; ftconstant in equation (5-56)
distance along meridional streamline, m ; ft
distance normal to streamline, m; ft
absolute pressure, N/m2; lb/ft 2
distance along arbitrary curve, m; ft
gas constant, J/(kg) (K) ; (ft) (lbf)/(lbm) (°R)
radius, m; ft
blade spacing, m; ft
absolute temperature, K; °Rtime, secstream function
absolute velocity, m/see; ft/see
relative velocity, m/see; ft/sec
mass flow rate, kg/sec; lb/sec
fluid absolute angle of inclination from axial direction in the
meridional plane, deg
fluid flow angle, relative to blades, out of the meridional plane
(in the tangential direction), deg
distance in direction of outer normal to cascade boundary, m; ftangular distance in direction of rotation, rad
density, kg/m3; lb/ft 3potential function
oangular velocity, rad/seeverrelaxation factor
Subscripts:
C
in
m
n
out
Pr
8
x
YZ
curvature
inlet
meridional component
component normal to streamlineoutlet
pressure surface
radial componentsuction surface
axial component
component in y-directioncomponent in z-direction
55
TURBINE DESIGN AND APPLICATION
0
°t1, 2,
3, 4
tangential component
mesh-point designations
56
CHAPTER6
Introductionto Boundary-LayerTheory
By WilliamD. McNally
As shown in chapter 2, the pressure ratio across a turbine provides a
certain amount of ideal energy that is available to the turbine for pro-ducing work. The portion of the ideal energy that is not converted towork is considered to be a loss. One of the more important and difficult
aspects of turbine design is the prediction of the losses.
Before losses can be predicted, it is necessary to understand their
causes. The primary cause of losses is the boundary layer that developson the blade and end-wall surfaces. Other losses occur because of shocks,
tip-clearance flows, disk friction (windage), flow incidence, and partial-
admission operation. This chapter gives an introduction to boundary-layer theory, which is used to calculate the parameters needed to estimate
viscous (friction) losses. 5Iethods for determining the basic viscous loss
and the associated trailing-edge and mixing losses are presented in thenext chapter.
NATURE OF BOUNDARY LAYER
When a real fluid (such as air) flows past a turbine blade at normal
velocities, the influence of viscosity on the flow is confined to a relatively
thin layer in the immediate neighborhood of the blade. This layer is called
the boundary layer. At the outer edge of this layer the flow is frictionless,and conditions there agree with those calculated with the use of ideal
(frictionless, nonviscous) flow assumptions. At the wall, on the other
hand, the velocity of the fluid is zero in all directions (no-slip condition).
57
TURBINE DESIGN AND APPLICATION
It is the frictional, or viscous, forces in this thin layer that reduce the fluid
velocity from its free-stream, frictionless value to zero at the wall.
A boundary layer on a turbine blade is illustrated in figure 6-1. The
boundary layer develops from a small finite thickness at the stagnation
point at the leading edge of the blade and grows along both the suction
and pressure surfaces. The initial portion of the boundary layer is always
laminar. In a laminar boundary layer, fluid layers parallel to the blade
surface slide over each other. Any minute local fluctuations in velocity are
sufficiently damped so that they have negligible influence on the smooth-
ness of the overall flow. The velocity at a point is either steady with time
or changes in some smooth way, as figure 6-2(a) indicates.
hIost flows being ducted to a turbine, or entering it from a combustor,are turbulent in nature. The fluctuating components of velocity have a
significant influence in this type of flow. With this overall flow, the
boundary layer on the blades cannot remain laminar for any great dis-tance. It usually passes through a transition region and becomes a tur-
bulent boundary layer. In the transition region, weak disturbances in the
flow are amplified, and this leads to the r'_ndom fluctuations in velocitythat are characteristic of turbulent flow. In the turbulent boundary layer,
as in turbulent flow, the velocity at any point oscillates in a random
fashion about a mean value, as figure 6-2(b) indicates.
Figure 6-1 also shows a separated region in the turbulent boundary
layer. Separation can likewise occur in the laminar boundary layer. When
a boundary layer separates, the fluid moves away from the blade surface.
The manner in which this happens is illustrated in figure 6-3. As the free-
stream velocity decreases along the rear portion of the suction surface ofa turbine blade, static pressure corresponding]y increases. This positive
r Transition r-Turbulent
Laminar _region / boundary_unda ry _ layer
er --._
' region -/
58
FmURE 6-1.--Boundary layer on blade.
vlINTRODUCTION TO BOUNDARY-LAYER THEORY
Steady
I Steady
V _'vW ,y-,,j-w','lr w- V v, ,v'W vv'
t t
I Unstea_
Unsteady _1_..o rV
Ill (b)
t t
(a) Laminar flow. (b) Turl)ulent flow.
Florae 6-2.--Variation of velocity with time at a point.
FmuaE 6-3.--Boundary-layer separation.
pressure gradient (adverse pressure gradient) retards the flow in the
boundary layer and causes it to lose energy. The flow in the boundarylayer can be retarded to such a degree that very close to the wall it moves
in a direction opposite to that of the mean flow passing the blade. This is
separation. The point at which the flow reverses itself is the separation
point. The laminar boundary layer at the leading edge of a turbine bladecan also separate and immediately reattach itself to the surface as a
turbulent boundary layer. This is illustratcd in figure 6-4.
Finally, it should be noted that both laminar and turbulent boundarylayers CaD be either incompressible or compressible, depending on the
level _)f the 5[ach number. Just as there are diffcrent equations to repre-
sent laminar and turbulent boundary-layer flow, there are different
equations for the incompressible and compressible variations of each.
59
TURBINE DESIGN AND APPLICATION
Turbulent
boundary
Sepa rat ion layer 7Laminar bubble 7 j...-__
boundary _¢_....__III _I 1_I II 1_' ' '
l.yer-,, ...., /.S_.liii'menl
point
FIGURE 6-4.--Laminar separation and reattaehment.
Boundary layers should be considered compressible if the free-stream
relative ._laeh number exceeds values of 0.3 to 0.4. The boundary-layerequations for these various cases arc derived and solution methods are
discussed in this chapter.
DERIVATION OF BOUNDARY-LAYER EQUATIONS
The general equations of motion of viscous fluids are called the Navier-
Stokes equations. In normal coordinate systems, there are three such
equations, one for each of the coordinate directions. The boundary-layer
equations can be derived from the Navier-Stokes equations. The Navier-
Stokes equations themselves can be derived by applying the law of con-
servation of momentum to a fluid element. This exercise is lengthy, and
will not be repeated here. References 1 and 2 both have the completederivation, in two somewhat different forms.
There are various forms of the Navier-Stokes equations, depending on
what assumptions are made during their derivation. The following equa-
tion represents the Navier-Stokes equations combined into vector formfor a compressible fluid with constant viscosity:
dU=gf_g Vp+g - V_u+ _ __V(V-u) (6-1)dt p p p
where
u general velocity vector, m/sec; ft/see
t time, sec
g conversion constant, 1; 32.17 (Ibm) (ft)/(lbf) (sec 2)f general body force acting on a unit mass of fluid, N/kg; lbf/lbm
6O
INTRODUCTION TO BOUNDARY-LAYER THEORY
p density, kg/mS; lbm/ft 3
p static pressure, N/m_; lbf/ft _t_ dynamic viscosity, (N) (sec)/m2; lbm/(ft) (sec)
In this equation, u represents a general velocity vector with components
u, v, and.w in the three coordinate directions x, y, and z, respectively.
u=ui+vj+wk (6-2)
where i, j, and k are the unit vectors in the three coordinate directions.
The total, or substantial, derivative of u is du/dt. In any of the coordinate
directions,
d 0 0 0 0
4t = 0-]+ u _x+V _y+W 0z (6-3)
In equation (6-1), the Laplacian operator V_ is applied to the vector urather than to a scalar function. If the term V2u is expanded into simple
vector quantities, equation (6-1) becomes
e--_=ge-'-qvp+"- v(v.u)-"- [vx(vxu)3+_ v(v.u) (6-4)dt p p p
Expressing the V operator in terms of gradients, curls, and divergences,
which may be more familiar to the reader, equation (6-4) becomes
du __-,qu=gf--p grad p+#-p grad(div u) p curl(curl u) +_ p grad(div u)
(6-5)
In order to derive the boundary-layer equations, equation (6-1) has
to be expanded into three scalar equations, one for each of thc coordinate
directions. The three resulting equations are
Ou Ou Ou Ou , g Op + _ /O_u O_ d_u\_+u_+v_+W0z:gJ,-_ ;(,_+_+_)
1_ 0 [Ou Ov Ow\
+-3;_(,_x+_+_) (6-6)
_+u _x+V_+_ _--g_,- +_ I,_+_+_)
1_ 0 /Ou Ov Ow\+_- _ (,_+_+_) (6-7)
61
TURBINE DESIGN AND APPLICATION
Ow Ow Ow Ow g Op . _ ['02w O:w 02w\
¥+u
1 a /Ou Ov Ow\+5"-
where f,, f_, and f, are the components of the body force f.
(6-8)
Laminar Incompressible Boundary Layer
In order to derive Prandtl's boundary-layer equations for laminar
incompressible flow, the following assumptions will t)e made:
(1) Viscosity is a constant. This has already been assumed in thewriting of the preceding equations.
(2) Flow is incompressible. Since for ineoinpressible flow the con-tinuity equation is
=(0u 0v 0w\V.u:div u \_xx+_y+_-z ) =0 (6.9)
the final terms in equations (6-6) to (6-8) can be eliminated.
(3) Flow is two-dimensional. This eliminates equation (6-8) from
consideration, as welt as all terms involving w or O/Oz in equations (6-6)and (6.7).
(4) Flow is steady. This eliminates O/Ot terms.
(5) Body forces are negligible in relation to inertia and viscous forces.
Thus, f_ andfu can be discarded from equations (66) and (6-7).
With these assumptions, the Navier-Stokes equations reduce to the
following two equations for the x- and y-direetions:
Ou _-v Ou g Op __". /02u 02u\ 'ox o. o (6.1o)
Ov Ov g Op __". /Oh, Oh,\u--+Vox Oy...... , Oy p _x_+_y :) (6-11)
Likewise, the continuity equation becomes
Ou Ov--+--=0 (6-12)Ox Oy
In order to make equations (6-10) to (6-12) suitable for the analysis of
boundary-layer flow, the equations are traditionally made dimensionless,
and an order-of-magnitude check is performed on the various terms to
show that some are negligible with respect to others. Figure 6-5 shows the
velocities and coordinate directions pertinent to the boundary layer.
62
INTRODUCTION TO BOUNDARY-LAYER THEORY
_6folt rra_l.g
FIGURE 6-5.--Boundary-layer velocities and dimensions.
The following dimensionless parameters are defined :
XX---
L(6-13a)
y=_L
(6-13b)
where
X
L
Y
U
U0V
ItU =-
Uo
V
V-Uo
gPP-
pUo_
pLUoRe _ --
dimensionless x-coordinate
characteristic length (in this case, the blade chord), m; ft
dimensionless y-coordinate
dimensionless velocity in x-direction
free-stream velocity upstream of blade, m/sec; ft/sec
dimensionless velocity in y-direction
(6--13c)
(6-13d)
(6-13e)
(6-13f)
63
TURB1NE DESIGN AND APPLICATION
P dimensionless pressureRe Reynolds number
From figure 6-5, we see that since x is proportional to L, X is of order 1.
And since y is proportional to the boundary-layer thickness 6i, u, Y is of
order 6i.u/L = _, a quantity much less than 1. Likewise, since u is of order
Uo, U= u/Uo is of order 1. And V=v/Uo is of order t, since velocities in
the y-direction in the boundary layer are much smaller than those in thex-direction.
In order to put equations (6-10) to (6-12) in terms of dimensionless
quantities, equations (6-10) and (6-11) are multiplied by L/Uo _, and
equation (6-12) is multiplied by L/Uo. The resulting dimensionlessequations are
OU OU OP 1 [02U O_U\
V_+Vo_= OX_et,_+_) (6-14)
OV OV OP 1 /02V 02V\
UUx+V --=OY --_ + Re _-Of2+_) (6-15)
OU OV
(6-16)
The order of magnitude of the various terms in these equations can now
be compared with each other. Since X and U are of order 1, and Y and Vare of order ,,
oU 1-= 1 (6-17a)0X 1
oU 1(6-17b)OY
OV-- _-= _ (6-17c)OX 1
OV--_-= 1 (6-17d)OY
02U 1- = 1 (6-17e)
OX _ 1.1
O_U 1 1-- --_-- = (6-17f)OY _ _.e d
64
INTRODUCTION TO BOUNDARY-LAYER THEORY
02V e_-- =_ (6-17g)
dX _ 1.1
02V _ 1(6-175)
OY _ e._
Furthermore, the change in P with respect to X is of the same order of
magnitude as the change of U with respect to X, so that OP/OX is oforder 1.
Relating these orders of magnitude to the terms in equations (6-14)to (6-16) yields
u OU v OU OP 1 /02U O2U\
(1)(1)+(e)(_)=--lT(e2)(l+_)
OV v OV OP 1 [O_V 02V\U _+ -_= OY F-_e[_+_) (6-19)
(1) (_) + (_) (1) = -_+ (e2) (_+!)
OU OV
_+0--_ =0 (6-20)
1+1
By examining equations (6-18) to (6-20), the following conclusionscan be reached:
(1) In boundary-layer theory, it is assumed that the viscous terms
1/Re[(O:U/OX 2) + (O_U/OY _) ] are of the same order of magnitude as the
inertia terms U(OU/OX)+V(OU/OY). For this to be true in equation
(6-18), 1�Re must be of order _2, since O_U/OY _ is much larger than
O_U/OX 2 and dominates the two terms in parentheses. Therefore, theReynolds number must be relatively large.
(2) In equation (6-19), with 1�Re of order _2 and with 02V/OY2
dominating O_V/OX 2, the terms are of order _. Therefore, unless OP/OY
is to dominate, it too must be of order _ or less. Therefore, OP/OY is much
smaller than OP/OX, and P can be considered a function of X alone. There-
fore, P=P(X) or p=p(x), and OP/OX=dP/dX or Op/Ox=dp/dx. This
allows us to assume that the pressure across the boundary layer in the
65
TURBINE DESIGN AND APPLICATION
y-direction is essentially constant. It can be assumed equal to the potential
flow pressure existing at the outside of the boundary layer.
(3) Since the first equation is of order 1, and the second equation is of
order _, the second equation can be neglected completely.
(4) In equation (6-18), O_U/OX _ can be neglected because it is so
small in comparison with O:U/OY _. This leaves the following dimension-
less equations :
OU OU dP 1 02U (6-21)dx
OU OV
(6-22)
These are Prandtl's boundary-layer equations in dimensionless form.
The boundary-layer equations in this form are useful in determining
the influence of the Reynolds number on the size of the boundary layer
for different fluids. From equation (6-21) we see that as Re increases in
magnitude, the viscous-force terms (1�Re) (02U/OY 2) will get smaller and
smaller. The boundary-layer thickness will correspondingly decrease. So,
as Re increases, 5],,, decreases. Furthermore, increasing Re corresponds to
decreasing viscosity if pLUo is constant. So, as a general rule, the thickness
of the boundary layer decreases as the viscosity decreases.
The boundary-layer equations can be put in terms of dimensional
variables by multiplying equation (6-21) by Uo2/L and equation (6-22)
by Uo/L. The resulting equations are
Ou Ou g dp tL 02uu -- + v .... { (6-23)
Ox Oy p dx p Oy2
Ou Ov
(6-24)
These are Prandtl's boundary-layer equations for two-dimensional,
]aminar, incompressible flow. Density and viscosity are assumed constant
and known. The pressure gradient along the blade surface, dp/dx, is also
known from an ideal-flow solution. The remaining unknowns are u and v,
and equations (6-23) and (6-24) are sufficient for their calculation.
It should bc noted that the boundary-layer equations are not valid in
the presence of shock waves (i.e., where instantaneous adverse pressure
gradients of large magnitude occur). Just as flow phenomena in theboundary layer depend on mainly the Reynolds number, conditions in a
shock wave depend on primarily the Mach number. Since the influence ofMach number is not included in the boundary-layer equations, they tell
us nothing about the interaction of shock waves and boundary layers.
66
INTRODUCTION TO BOUNDARY-T..k-,r.rt TIIEORY
The boundary-layer equations are not completely reliable as separationis approached. One of the assumptions used in their derivation is that the
velocity v is much smaller than _. Very close to the separation point, theboundary layer grows rapidly, and v begins to be of the same order as u.
Nonetheless, the boundary-layer equations are generally used in cal-
culations right up to the separation point, since the region where V is
significant is very small, and little error in the location of the point ofseparation is incurred. However, these equations should not be used for
detailed calculations in the neighborhood of a separated flow region.
The Navier-Stokes equations (6-6) to (6-8) used in the development
of the boundary-layer equations were derived for an orthogonal systemof coordinates in which the radius of curvature of each of the coordinate
axes is quite large (i.e., where curvature effects are negligible). The
question arises as to how the boundary-layer equations would change for
flow over a curved wall. If a curvilinear orthogonal coordinate system(fig. 6-6) is introduced wherein the x-axis is in the direction of the curved
wall and the y-axis is normal to it, a new set of Navier-Stokes equationscan be derived for flow in such a system. These equations are given in
reference 1. The terms in the equations are very dependent on _hc radius
of curvature r at a position x along the blade surface. The relative orders
of magnitude of the individual terms can be estimated in the same manner
as was done previously. With the assumption that the boundary-layer
thickness is small compared with the radius of curvature of the wall, and
for the case where no large variations in curvature occur, so that dr/dx.-_ l,the same boundary-layer equations result as were obtained for flat walls.
Therefore, the flat-plate boundary-layer equations may be applied to
curved walls as well, provided there are no large variations in curvature,such as would occur near sharp edges.
y
X X
Y
FIGURE 6--6.--Curvilinear coordinate system on a blade.
67
TURBINE DESIGN AND APPLICATION
Laminar Compressible Boundary Layer
An order-of-magnitude analysis can also be performed to derive the
equations for a compressible boundary layer. In the incompressible case,
viscosity and density were assumed constant, temperature variations were
neglected, and the energy equation was not used. For the compressible
case, density is no longer constant, viscosity is considered a function of
temperature, the equation of state is used to relate pressure and density to
temperature, and, if the process is not isothermal, some form of the energy
equation is required. The boundary-layer equations for compressible,nonisothermal, variable-viscosity flow will involve three parameters which
can be related to temperature. These are viscosity, specific heat, and
thermal conductivity.
There are several relations for viscosity as a function of temperature.
The most common is probably Suthcrland's relation (ref. l)
tt (T_3/2To+S_0 = \To/ T+S (6-25)
where
T
ToS
dynamic viscosity at the reference temperature To, (N)(sec)/m2;lbm/(ft) (sec)
absolute static temperature, Is:; °R
reference temperature, K; °R
a constant, K; °R (for air, S=llO K or 198 ° R)
A less complicated, but also less accurate, temperature-viscosity relation
is the power law
- = 0.5 <_ < 1.0 (6-26)tto
where _0 is a constant. For air, o_is approximately 0.65.
Specific heat and thermal conductivity can be related to temperature
by least-squares polynomial-curve fits for the particular gas and tem-
perature range involved. With these variables related to temperature,the unknowns in the compressible-boundary-layer problem reduce to u,
v, p, and T. The four equations relating these variables will be the con-
tinuity equation, one component of the momentum equation, the state
equation, and the energy equation.
The order-of-magnitude analysis of the continuity and momentum
(Navier-Stokes) equations for compressible flow is almost identical to
that for incompressible flow. For compressible flow with nonconstant
viscosity, the equations analogous to (6-10) to (6-12) are the following:
68
INTRODUCTION TO BOUNDARY-LAYER THEORY
au au ap, o r O_ 2 /0_ ov\l 0 [ /Ou Ov\l
(6-27)
Ov Ov Op, 0 F Ov 2 /au Ov\l O [ /Or au\l
(6-28)
a(pu) __a(pv)=0 (6-29)Ox Oy
If an order-of-magnitude analysis is performed on these equations similar
to that for the incompressible-flow equations, the following boundary-
layer equations result:
ou ,tp, o (6-3o)
a(pu) __a(pv) =0 (6-31)Ox Oy
The equation of state is also required for the solution of compressible
boundary-layer flow. The state equation is
p= pRT (6-32)
where R is the gas constant, in J/(kg) (K) or (ft) (lbf)/(lbm) (°R).
The final equation required besides the continuity equation, the
momentum equation, and the equation of state, is the energy equation.
The energy equation for a compressible boundary layer is derived from
the energy equation for a perfect gas by means of another order-of-
magnitude check. The folh)wing is the energy equation for compressible,two-dimensional steady flow of a perfect gas, written in full :
pc,, (uOT aT\ uOp . v Op . O /, aT\ O (kOT_+ u ,_
where
Cp
Jk
and
(6-33)
specific heat at constant pressure, J/(kg)(K); Btu/(lbm)(°R)conversion constant, 1;778 (ft) (lbf)/Btu
thermal conductivity, W/(m) (K); Btu/(see) (ft) (°R)
F/auV /avV1 /Or auV 2/au avV(6-34)
69
TURBINE DESIGN AND APPLICATION
If an order-of-magnitude check is performed on tile above equations, thefollowing boundary-layer energy equation results:
pcp (u OT OT\ udp 0 /k OT\ u (Ou_ 2 (6-35)
Equations (6-30), (6-31), (6-32), and (6-35) are the laminar boundary-layer equations for nonisothermal, two-dimensional, compressible flow ofa gas obeying the ideal gas law.
Turbulent Boundary-Layer Solution Methods
It is desirable to have a turbulent boundary layer over the major portionof turbine blades. If the boundary layer is not turbulent, separation will
probably occur on the blades, with a resulting decrease in their per-
formance. Turbulent flow has irregular fluctuations (mixing or eddymotions) _uperimposed on the main fluid motion (see fig. 6-2). These
fluctuations are so complex that closed-form solutions are not feasible at
present. Yet, the mixing motion is very important, since the stresses in
the fluid due to fluctuating components of velocity are often of greatermagnitude than those due to the mean motion.
There are two approaches to the solution of turbulent boundary-layerflow. The first is the exact solution of the time-dependent, three-dimen-
sional, Navier-Stokes equations. The three-dimensional equations are
required, since two-dimensional calculations could newer represent thestretching of eddies, which is a prime mechanism of turbulent flow. How-
ever, even the largest computers available at the present time cannot.
handle three-dimensional solutions of these equations on a small enoughmesh to represent the fluctuating components of velocity of turbulentflOW.
The second approach is to write the equations of continuity, momentum,
and energy in terms of mean and fluctuating components of pressure,
density, temperature, and velocity. In this approach, the time averageof the u component of velocity, for example, is denoted by zZ and the
w_locity of fluctuation by u'. So the velocities, density, pressure, andtemperature are writ ten as follows:
u = a+ u' (6-36a)
v=_+v' (6-36b)
P= P+ P' (6-36c)
P=P+P' (6-36d)
T = 7"+ T' (6-36e)
The fluctuations in viscosity, thermal conductivity, and specific heat are
70
INTRODUCTION TO BOUNDARY-LAYER THEORY
negligible and are not considered. So these three parameters are calculated
as functions of the time-averaged value of temperature.
If the flow properties listed in equations (6-36) are substituted into the
continuity, momentum, and energy equations for incompressible and
compressible flow, a new set of stress terms arises in thees equations.These are called the "apparent" turbulent stresses, or Reynolds stresses.
They are of the form pu'----iand ou'v---_, where u'v' is the average over time of
the product of u' and v'. These new terms in the equations add additionalunknowns to the boundary-layer problem for which additional equations
are not presently available. For this reason, empirical expressions or
approximations are substituted for the Reynolds stress terms before th_
turbulent boundary-layer equations can be solved.
Turbulent Incompressible Boundary Layer
Substituting the relations of equations (6-36) into equations (6-10),
(6-11), and (6-12), and then performing an order-of-magnitude analysis
yields the following equations for turbulent, incompressible, boundary-
layer flow:
oa oa dp. 0 ( oa .__.,'_aft _x+p_ _ = --g _x-t-_y \t* -Z---Puoy v /I (6-37)
0_+0__ = o (6-38)Ox Oy
These equations are analogous to equations (6-23) add (6-24) for laminar
flow. Notice, however, the presence of the Reynolds stress term ill the
momentum equation. This adds a new unknown (u'v') to the original two
(_ and _), thereby making three unknowns with only two equations.
Turbulent Compressible Boundary Layer
Substituting the relations of equations (6-36) into equations (6-27),
(6-28), (6-29), (6-32), and (6-33) and then performing an order-of-
magnitude analysis yields the following equations for turbulent, com-
pressible, boundary-layer flow:
O(_(t) __O(li_) +O(p'v'_._))=0 (6-39)Ox Oy Oy
-g v} (6-4o)
= _R_ (6-41)
71
TURBINE DESIGN AND APPLICATION
Y� y Jdx+Oy Oy/
1 0 # k O_ 0 (6-42)
where T_ is the absolute total temperature, in K or °R, and is defined as
T,= T+ <.43)
We have now derived the basic boundary-layer equations for two-
dimensional, laminar and turbulent, incompressible and compressible
boundary-layer flow. We should note at this time that this is really onlythe starting point as far as boundary-layer solutions are concerned.
These equations are only the basis for the many, many methods whichpresently exist for obtaining boundary-layer solutions under variouscircumstances.
SOLUTION OF BOUNDARY-LAYER EQUATIONS
After velocity profiles are discussed and the important boundary-layerparameters defined, some of the solution methods will be discussed.
hicluded will be the flat-plate, incompressible solution, as well as com-pressible methods.
Velocity Profiles
One of the principal results obtained from most boundary-layer solu-
tions is a description of the velocity profile in the boundary layer along
the blade surface (fig. 6-7). The velocity profile describes mathematically
Outeredge , Uej_
ofboundary
72
FI(]UaE 6-7.--Boundary-layer velocity profiles.
INTRODUCTION TO BOUNDARY-LAYER THEORY
Outeredgeolboundarylayer_
/ I
(a) Laminar (b) Turl)ulent
profile, profile.
FmvlcE 6-8.--Laminar and turbulent velocity profiles.
the dimensionless veh)city u/u. as a function of the dimensionless distance
Y/SI,, from the blade. The velocity u is the velocity in the boundary layerat a distance y from the surface, and the velocity ue is the external free-
stream velocity at a distance equal to the boundary-layer thickness,5I,,,,, from the surface. Alternately, 5iu,z is often defined as that distance
from the blade where the velocity differs by 1 percent from the externalvelocity, u,.
Velocity profiles for laminar flow (fig. 6-8(a)) tend to be parabolic in
shape, while those for turbulent flow are blunted (fig. 6-8(b)). A com-
monly used mathematical expression for u/u, in laminar flow is that
originated by Pohlhausen (see ref. 1) :
?_e _full \_s,,./
The constants a, b, c, and d are defined in terms of a dimensionless shape
parameter
X- _'' du, (6-45)t_ dx
where
a=2+_ (6-46a)
b= --- (6--46b)2
k
c= -2+_ (6-46c)
73
TURBINE DESIGN AND APPLICATION
Shapeparameter.X
- 7.052yJ_Ji
/o/JSJ
• 2 S "_]6'_
I I I I l.2 .4 .6 .8 1.0
Fraction of boundary-layer height, y/6full
FIGURE 6-9.--Laminar velocity profiles.
xd = i - - (6-46d)
6
Velocity profiles for various values of X are shown in figure 6-9.
Velocity profiles for turbulent flow are often represented by the power
law
,'_( 7J _},fn (6-47)Ue- \(_[. ll/
Pipe-flow experiments show that the exponent n is a mild function of the
Reynolds number and varies from 4 up to about 10. The value of n= 7is most appropriate for boundary-layer flow on a fiat plate. The exponentn can be related to other boundary-layer parameters, namely the dis-
placement thickness _ and the momentum thickness 0, which are describedin the next section.
Definitions of Important Boundary-Layer Parameters
Solutions of the two-dimensional boundary-layer equations are most
often obtained in terms of three important parameters. These are the dis-
placement thickness _, the momentum thickness 0, and the form factor H.In order to define these parameters, it is necessary to first define the
thickness of the boundary layer, _l_zz. The definition of boundary-layer
thickness is rather arbitrary, since transition from the velocity inside the
boundary layer to that outside it tak('s place asymptotically. This is of
little importance, however, because the velocity in the boundary layerattains a value which is very ch)se to the external velocity at a small
distance from the wall. It is possible to define the boundary-layer thick-
74
-7-_ull
llllllllllJ
INTRODUCTION TO BOUNDARY-LAYER THEORY
U e
,-(ue - u)
(a)'fill, IIIII,
U e=-
Me
IIIIIIIIIIIIII
(b)IIII111111
(a) Actual velocity profile. (b) Equivalent profileequal mass flow.
FIGURE 6-10.--l)isplacement thickness of a houndary layer.
for
ness as that distance from the blade where the velocity differs by 1 percentfrom the external velocity ue.
The displacement thickness _, for compressible boundary-layer flow,
can be defined with the help of figure 6-10. As seen from figure 6-10(a),the decrease in mass flow within the boundary layer due to the influence offriction is given by
[ass defect =/_:_i.- (p_u_-- pu) dy (6-48)y-.0
where pe is the density, in kg/m 3or lbm/ft ._,in the free stream outside of
the boundary layer. This integrated mass defect can be represented by a
distance _, the displacement thickness, as shown in figure 6-10(b). It is
the distance by which the external potential field of flow is displaced
outward as a consequence of the decrease in velocity in the boundarylayer.
As figure 6-10 shows, the distance _ can be defined by the equation
fy=51ult
p_Le_= ] _0 (p,u_-pu) dy (6--49)
Solving for $ gives
a=p--uu, j,_0 (p.u,-pu) dr= pu
The displacement thickness for incompressible flow reduces to
5= (_,-u) dy= dy (6-51)Ue v-o w-o \ _te/
The loss of momentum in the boundary layer due to the presence offriction is given by
75
TURBINE DESIGN AND APPLICATION
._Iomentum defect = r]_ =_l,,. pu (Ue-- U) dy (6-52)J_ y=0
This momentum defect from the momentum of purcqy potential flow
can be represented by a distance 0, defined by the equation
Y=_fuHp_ue20= p_ ( _te-- u) dy (6-53)"y--O
Solving for 0 in this equation gives the definition of the momentum thick-
ness for compressible boundary layers as
0=-- pu(_--_) ely= pu 1 -u_- dy (6-54)Pe _le2 _ y=O _ 1/=0 pc?le %k _e/
The momentum thickness for incompressible flow reduces to
0=-_ 2 -o .(,l_--u) dy=] -- 1-"-y--O ;_le \ He/
The form factor H for both compressible and incompressible flow is
defined as the ratio of displacement thickness to momentum thickness:
5H = - (6-56)
0
There are many other boundary-layer parameters besides 5, 0, and H for
two-dimensional, and especially for three-dimensional, boundary layers.
These three, however, are the principal parameters used in general
boundary-layer studies.
Physical Interpretation of Separation
When separation of flow from a blade or a casing occurs, some of theretarded fluid in the boundary layer is transported away from the surface
toward the main stream. When a region with an adverse pressure gradient
exists along a surface, the retarded fuid particles cannot, in general,
penetrate too far into the region of increased pressure because of their
small kinetic energy. Thus, the boundary layer is deflected away from thesurface and moves into the main stream. In general, the fluid particles
behind the point of separation follow the pressure gradient and move in a
direction opposite to the external stream. The point of separation isdefined as the limit between forward and reverse flow in the layer in the
immediate neighborhood of the wall. At separation,
Figure 6-11 illustrates separation occurring along a surface.
76
INTRODUCTION TO BOUNDARY-LAYFR THEOR_
>0 _ -0 (__u_
- "Separation ...._x\\\'_ ' '_ _ y'O < 0
point
FlG(_'l¢l.: 6--1 l.--Velocity gradients as flow undergoes separation.
By examining Prandtl's boundary-layer equations and considering the
relation between pressure gradient dp/dx and velocity distribution u(y),
it is possible, to infer that separation in a steady flow will occur only in
the presence of an adverse pressure gradient (i.e., decelerated flow),
dp/dx>O. From equation (6-23), with the boundary conditions at the
surface being u = v = 0, we have
(O2u_ dp_' \og_/,=o = g (Ix (6--58)
We can now relate velocity profiles to Ou/Oy, 02u/Og 2, and finally to
dp/dx through equation (6-58). The e(luation indicates that in the
immediate neighborhood of the wall, the curvature of the velocity profile,
02u/Oy _, depends only on the pressure gradient, dp/dx, and the curvature
of the velocity profile at the wall changes its sign with the pressure
gradient.
Figure 6-12(a) shows a velocity profih_ that would exist in a boundary
layer subjected to a decreasing pressure. For such a profile, figure 6-12 (b)
(b) Ic)
(a) Velocity (b) Velocity (c) Velocity-
profile, gradient, profileeurv._ture.
FIGURE 6-12.--Velocity distribution in a boundary layer with pressure decrease.
77
TURBINE DESIGN AND APPLICATION
indicates that Ou/Oy is positive for all y and decreases as y increases.
Furthermore, figure 6-12(c) indicates that O_u/Oy _, which is the slope of
Ou/Oy, is negative for all y. From equation (6-58), we know that negativeO_u/Oy 2 corresponds to negative dp/dx. Consequently, a boundary layer
subjected to a decreasing pressure (negative dp/dx) will have velocityprofiles which are not indicative of impending separation (the form of
fig. 6-12 (a)).
Figure 6-13(a) shows a profile which would cxist in a boundary layer
with decelerated flow due to an increasing pressure (adverse pressuregradient). Here, figure 6-13(b) indicates that Ou/Oy has a positive slope
near the blade surface; that is, O*u/Oy 2 is positive (fig. 6-13(c)). This
corresponds to positive dp/dx. However, since in all cases O_u/Oy _ mustbe less than zero at some distance from the surface, there must exist a
point for which O_u/Oy_= O. This is a point of inflection of the boundary-
layer velocity profile. It follows that in a region of retarded potential flow
yPoint of
__"_ _ _Inflect;on
////,
(a) Velocity (t)) Velocity (c) Velocity-
profile, gradient, profile curvature.
FmuR_ 6-13.--Velocity distribution ill a boundary layer with pressure increase.
-_$. tacjnat ion
_- Adverse gradient
FIGURE 6-14.--Pressure distribution on a turbine blade.
78
INTRODUCTION TO BOUNDARY-LAYER THEORY
(positive dp/dx), the velocity profile in the boundary layer will have a
point of inflection. Since the velocity profile at the point of separation
(with Ou/Oy = 0 at the surface) must have a point of inflection, it follows
that, with the assumptions used in deriving equation (6-23), separation
can occur only when the potential flow is retarded (i.e., in regions ofaxiverse pressure gradient).
Figure 6-14 indicates a typical pressure distribution on the surface of a
turbine blade. The danger zone, as far as separation is concerned, is
readily seen to be the rear portion of the suction surface, where the majorpart of the blade diffusion is taking place.
Laminar Incompressible Boundary Layer on a Flat Plate
Prandtl's boundary-layer theory was first reported in 1904 in Germany.It was later translated and published in 1928 as an NACA Technical
Memorandum (ref. 3). The first mathematical solution of Prandtl's
equations to be published was the flat-plate solution of Blasius in 1908.
This German work was also later translated by NACA (ref. 4).
On a fiat plate with steady flow at zero incidence, the velocity from the
potential solution is constant. Therefore, p (x) is constant and dp/dx = O.The boundary-layer equations, therefore, reduce to
Ou Ou 02uu _+v _ (6-59)
ay Oy2
where v is the kinematic viscosity u/p, in m2/sec or ft2/sec, and
Ou c_v
(6-60)
The following are the boundary conditions:
u=v=O at y=O
u=u, at y=(6-61)
With the use of a stream function _h, Blasius transformed the partial
differential equation (6-59) into the following ordinary differentialequation:
fdd_2+2 d3f 0dy---_=
whel;e f is a normalized stream function
(6--62)
f(') - %/v-x_ (6-63)
79
TURBINE DESIGN AND APPLICATION
which depends on the dimensionless y-coordinate, 7, where
Y(6-64)
This equation has the following boundary conditions:
f=d_f =0 at 7=0ay
df--=1 at 7=dy
(6--65)
Equation (6-62) cannot be solved exactly. Blasius obtained an approxi-
mate solution in the form of a power series expansion about 7=0 and an
asymptotic expansion for 77= _, the two solutions being joined at a suit-
able point. More recently, Howarth (ref. 5) solved the Blasius equation
(6-62) with a high degree of accuracy, and provided tabular values for f,
df/dy, and d:f/dy 2as functions of 7. Since df/dy = u/ue, the solution gives
the velocity profile of figure 6-15. This profile possesses a very small
curvature at the wall and turns rather abruptly further from it in order
- 5
_3
c
2
e
e-
_- 0 .2 .4 .6 .8 1.0
Boundary-layer velocity ratio, u/u e
FIGURE 6-15.--Blasius-Howarth velocity profile for flow on a fiat plate.
8O
INTRODUCTION TO BOUNDARY-LAYER THEORY
to reach the asymptotic value. At the wall itself, the curve has a point ofinflection, since for y = O, 02u/Oy 2= O.
From the order-of-magnitude analysis performed to obtain Prandtl's
boundary-layer equations, we had the relation
1_ _8_zl_w- (6-66)
For a semi-infinite flat plate, the Reynolds number can be expressed as
U_
Re= = -- (6-67)p
In order to make equation (6-66) dimensionally correct, we can say
8_ull ¢¢ 1X2 Re= (6-68)
or
_j':z__ v_ (6-69)
The constant of proportionality can be obtained from Howarth's numerical
solution and is equal to 5. So, for a semi-infinite flat plate at zero incidence
in laminar flow, we obtain the useful relation for the boundary-layerthickness
_/_z_= 5.0 _ (6-70)
With the use of Howarth's solution to the Blasius equations, the follow-
ing relations for other important boundary-layer parameters for laminarflow on a flat plate can also be obtained:
8=1.72 _ (6-71)
0=0.664 v_, (6-72)
gr_ 0.332= 0.332 _l"
1_3_8
D = b%/_pluo _ (6-74)g
81
TURBINE DESIGN AND APPLICATION
where
1.328CI= 1.328 _]-_
"uet(6-75)
Tw
D
b
l
CfRez
shear stress on the surface, N/m:; lbf/W
total drag on both sides of fiat plate, N; lbf
width of flat plate, m; ft
length of fiat plate, m ; ftdimensionless drag or skin friction coefficient for fiat plate
Reynolds number based on plate length l
It should be noted that all of these relations are valid only for laminar
flow; that is, they are valid only where Re_< 106, a value that is indicativeof laminar flow over the entire length of the plate. For Ret> 106, transition
to turbulent boundary layer will probably occur, and the expressions in
equations (6-71) to (6-75) will be valid only from the leading edge of the
plate to the transition point. If transition to turbulent boundary layer does
occur, then the drag will be larger than that calculated by equation
(6-74).
Integral Methods for Solving the
Laminar-Boundary-Layer Equations
The two principal means of solving the laminar-boundary-layer equa-
tions are by integral methods and by finite-difference methods. Both
means provide approximate solutions, since exact solutions are extremelycumbersome.
Integral methods are based on yon K£rm_in's momentum integralformula. Von K_rm_.n's original work was published in 1912 in Germany
and was later translated by NACA (ref. 6). Von Kgrmgn realized that it
was not necessary to satisfy the boundary-layer equations for every fluid
particle. Instead, he satisfied the boundary-layer equations close to thewall and in the region where external flow is approached by satisfying the
boundary conditions. In the remaining region of fluid in the boundary
layer, only a mean over the differential equation is satisfied. Such a meanis obtained from the momentum equation (eq. (6-23) or (6-30)) by
integration over the boundary-layer thickness. If equations (6-23) and
(6-30) are integrated from y =0 to y = _1,,u, and if the definitions of dis-
placement thickness (eq. (6-50)) and momentum thickness (eq. (6-54))are introduced, the following equations result. For laminar, incom-
pressible flow,
dO du_ gr_ (6-76)
82
INTRODUCTION TO BOUNDARY-LAYER THEORY
For laminar, compressible flow,
dO dub grw (6--77)u_ dx -{-(20+_- Me20)ue -_x = p_
where the subscript e denotes conditions at the outer edge of the boundary
layer.
Equation (6-76) or (6-77) leads to an ordinary differential equationfor the boundary-layer thickness, provided that a suitable form is as-
sumed for the velocity profile, u/u,. This allows us to calculate the dis-
placement thickness, 3, the momentum thickness, 0, and the shearingstress at the wall, rw. Pohlhausen was the first to use equation (6-76) to
obtain a solution for incompressible boundary layers. His work was
published in 1921 (refs. 7 and 1). The velocity profile assumed by
Pohlhausen was discussed earlier in this chapter, under "Velocity Pr<)-
files." Although Pohlhausen's solution is probably the simplest, it is known
to giw, poor results in regions of rising pressure. As a result, various authors
have tried to improve and extend his method by assuming different
families of velocity distributions.
A famous work among those which followed Pohlhausen's was that of
Thwaites (ref. 8). Thwaites collected and compared all known velocity
distributions from exact and approximate solutions for laminar incom-
pressible flow. Thwaites' method does not require the solution of ordinary
differential equations. He relates the wall shear, its derivative at the wall,
and the form factor to one another without specifying a type of velocity
profile. To do this, nondimensional forms of these quantities were defined
and evaluated with the use of exact solutions for the laminar boundary
layer. It developed that a nearly universal relation existed among these
quantities for favorable pressure gradients. For adverse gradients,
Thwaites selected a single representative relation. A unique correlation
was chosen that reduced the solution of an incompressible problem to theevaluation of a single integral. Thwaites' method was extended to com-
pressible fluids by Rott and Crabtree (ref. 9). They recognized that when
heat transfer is negligible, and the Prandtl number is equal to 1, a trans-
formation proposed by Stewartson (ref. 10) could be used to relate
compressible to incompressible boundary-layer solutions.
One of the best integral methods to appear to date for the solution oflaminar boundary layers is that of Cohen and Reshotko (refs. 11 and 12).
Their method applies to compressible or incompressible flow over two-
dimensional or axially symmetric surfaces. It handles arbitrary free-
stream pressure distribution and performs well in areas of adverse pres-
sure gradient. A surface temperature level may be specified, and heattransfer is calculated. Cohen and Reshotko's method is based on Thwaites'
correlation concept. Stewartson's transformation (ref. 10) is first applied
83
TURBINE DESIGN AND APPLICATION
to Prandtl's equations. The resulting nonlinear, first-order differential
equations are expressed in terms of dimensionless parameters related to
the wall shear, the surface heat transfer, and the transformed free-stream
velocity. Then Thwaites' concept of a unique interdependence of these
parameters is assumed. The evaluation of these quantities is then carried
out by utilizing the exact solutions of reference 11. With the resulting
relations, methods are derived for the calculation of all the important
boundary-layer parameters. In 1960, Luxton and Young published a
method (ref. 13) which is as general as Cohen and Reshotko's, but whichallows the Prandtl number to have values slightly different from 1.
Finite-Difference Methods for Solving the
Laminar-Boundary-Layer Equations
Finite-difference methods for solving the boundary-layer equations
have recently come into prominence because of the development of
digital computers. Smith and Clutter have done a considerable amount of
work in developing this technique (rcfs. 14 and 15). Another recentreference of interest is that of Krause (ref. 16). These methods give very
good results with relatively short running times on the computer.
Eddy-Viscosity and Mixing-Length Concepts in
Turbulent Boundary-Layer Flow
Bcfore referencing any of the current, methods for solving turbulent
boundary-layer flow, the concepts of "eddy viscosity" and "mixing
length" should be discussed. These approximation concepts have beenused in many of the methods developed to date to relate the Reynolds
stresses produced by the mixing motion to the mean values of velocity
components. By this means, the Reynolds stresses are given a mathe-matical form which, upon substitution into the governing equations,
leads to differential equations containing only mean values of density,
velocity, and pressure. These transformed differential equations con-stitute the starting point for the calculation of the mean boundary-layer
flow.
Boussinesq first worked on this problem in 1877. In analogy with thecoefficient of viscosity in Stokes' law for laminar flow
r, - (6-78)g 0Y
where rt is the laminar shear stress, in N/m 2 or lbf/ft 2, he introduced a
mixing coefficient, A,, for the Reynolds stress in turbulent flow by putting
A, 0_ir, - (6-79)
g Oy
84
INTRODUCTION TO BOUNDARY-LAYER THEORY
where r, is the turbulent shear stress, in N/m 2or lbf/ft _. In 1880, Reynolds
introduced the concept of eddy, or virtual, viscosity, _, where
Ar
= -- (6-s0)P
Thus, the eddy viscosity is analogous to the kinematic viscosity _=t_/p.Turbulent stress can then be expressed as
p O_ -- p u_v--_T, =- _ --= -- (6-81)g Oy g
With the use of this concept, terms in equations (6-37) and (6-40) such as
Y /
can be written as
A similar concept can be applied to the energy equation where an eddy,
or a virtual, conductivity can be defined. The difficulty with applying the
eddy-viscosity method is that A, and hence e depend on velocity. It is,therefore, necessary to find empirical relations between these coefficientsand the mean velocity.
In 1925, Prandtl introduced a completely different approximation for
the Reynolds stresses. His argument is called Prandtl's mixing-length
hypothesis, since the mixing length is somewhat analogous to the meanfree path in the kinetic theory of gases. The main difference is that kinetic
theory concerns itself with the microscopic motion of particles, whereas
Prandtl's concept deals with the macroscopic motion of large clusters of
fluid particles. Deriving Prandtl's expression for shear stress requires agood deal of discussion of his physical model of turbulent flow, all of which
is contained in reference I. His final expression is
r, =- l2 = _o_ u'v' (6-82)g g
where l is the mixing length, in m or ft.
On comparing Prandtl's expression (eq. (6-82)) with that of Boussinesq
(eq. (6-81)), it appears that little has been gained. The unknown eddyviscosity _of the first expression has merely been replaced by the unknown
mixing length l of the second expression. However, Prandtl's equation(6-82) is generally more suitable for the calculation of turbulent motion
than is equation (6-81). Turbulent drag is roughly proportional to the
85
TURBINE DESIGN AND APPLICATION
square of velocity, and the same result is obtained from (6-82) if themixing length is assumed to be independent of the magnitude of velocity.
So, mixing length is a purcly local function, although we cannot say it is a
property of the fluid. It is far simpler to make assumptions about themixing length 1 than about the eddy viscosity e, and this constitutes the
superiority of Prandtl's expression over that of Boussinesq.
Integral Methods for Solving theTurbulent Boundary-Layer Equations
Just as with the laminar-boundary-layer equations, there are both
integral methods and finite-difference methods for solving the turbulent
boundary-layer equations. Both of these provide approximate solutions,since exact solutions for turbulent flow are now impossible.
Gruschwitz was the first to propose a method for solving the equations
for an incompressible turbulent boundary layer. His work was published
in Germany in 1931. A rash of works followed, most of them making
improvements to the calculational technique and empirical data used byGruschwitz. Ludwieg and Tillmann, whose work was published in Ger-
many in 1949 and was translated by NACA in 1950 (ref. 17), proposed an
empirical relation for the skin-friction term in the momentum integral
equation. This relation is still used in many current methods. Stewartson'stransformations (ref. 10) are likewise used in many methods for solving
the turbulent-boundary-layer equations.
_[askell, in 1951 (ref. 18), proposed an improved method for incom-
pressible turbulent boundary layers. He replaced the momentum equation
by an empirically determined approximation which is directly integrableand thus determines the momentum thickness. A profile parameter is
obtained from an empirical auxiliary differential equation. The Ludwieg-Tillmann skin-friction formula is used to calculate the skin-friction dis-
tribution and to determine a separation point for flows with adverse
pressure gradient.Truckenbrodt, whose work was published in Germany in 1952 and was
translated by NACA in 1955 (ref. 19), proposed solutions for both laminar
and turbulent incompressible boundary-layer flows. The method is simple
and, like Maskell's method, does not use the momentum integral equa-tion. It applies to both two-dimensional and rotationally symmetrical
flows. Because of its simplicity and relatively accurate results, Trucken-
brodt's method is still used for incompressible turbulent boundary layers.
Compressible turbulent boundary layers were first treated adequately
with the use of integral methods by Reshotko and Tucker in 1957 (ref.
20). Prior to their work, the Kgrmgn momentum integral equation hadbeen utilized with an assumed boundary-layer velocity profile, usually the
power law, and one of several empirical skin-friction relations. When
86
INTRODUCTION TO BOUNDARY-LAYER THEORY
pressure gradient was present, an auxiliary equation, usually the moment-
of-momentum equation, was used. (This equation is obtained by mul-
tiplying the integrand of the momentum integral equation by a distance
normal to the surface and then integrating with respect to that distance.)
The momentum integral equation and the auxiliary equation were thensolved simultaneously.
Reshotko and Tuckcr's method, applicable to compressible flow withheat transfer and pressure gradient, also uses the momentum and moment-
of-momentum integral equations. These are expressed in incompressibleform and are uncouplcd with the use of Stewartson's transformation
(ref. 10) and the results of Maskell (ref. 18). The Ludwicg-Tillmann
skin-friction relation is used in a form suitable for compressible flow with
heat transfer through application of Eckcrt's rcference-enthalpy concept
(rcf. 21). An approximation for the shear-stress distribution through the
boundary layer and the power-law velocity profile arc used to simplify the
moment-of-momentum equation. Separation is located as the point where
the skin friction, when extrapolated, becomes zero. This method, until
several years ago, was the best available for compressible turbulent
boundary layers. It is still widely used in many computer programs today.
One of the best integral methods available today for compressible
turbulent boundary layers is that of Sasman and Cresci (ref. 22). It issimply an extension of the Reshotko-Tucker method. It uses somewhat
the same analysis, but no attempt is made to uncouple the momentum
and moment-of-momentum integral equations. These equations arc solved
simultaneously after introduction of boundary-layer shear-stress dis-
tributions obtained from recent numerical results of equilibrium turbulentboundary-layer analysis. The Sasman-Cresci analysis is better than that of
Reshotko-Tucker at predicting separation in regions of adverse pressuregradient. McNally (ref. 23) has developed a computer program based on
the Cohen-Reshotko (refs. 11 and 12) and Sasman-Cresci (ref. 22) tech-
niques. An additional source of information on compressible turbulent
boundary-layer analysis is the work of Herring and Mellor (ref. 24).
Finlte-Difference Methods for Solving the
Turbulent Boundary-Layer Equations
Finite-difference methods for solving the turbulent boundary-layerequations have recently begun to appear. Cebeci and Smith have done a
large portion of this work to date (refs. 25, 26, and 27). Bradshaw, Ferriss,
and Atwcll have also developed methods for the turbulent boundary layer
(refs. 28 and 29) based on the use of the turbulent energy equation.
Patankar and Spalding have developed still another method for handling
the turbulent boundary-layer equations (refs. 30 and 31). A great deal of
work is going on in this field at the present time, and no method is yet
87
TURBINE DESIGN AND APPLICATION
clearly superior to any of the others. Two relatively recent publications
(refs. 32 and 33) compare many of the most prominent methods, both
integral and finite difference, for solving the turbulent boundary layer.
CONCLUDING REMARKS
The selection of a method of solution suitable to a particular boundary-
layer problem requires some familiarity with the various methods avail-able. This can be achieved by studying some of the more recent references
that have been mentioned herein. The present discussion of the methods
of solution has been intended to show the historical development of
solution techniques, the variety of methods available, and the complexity
of the whole boundary-layer problem, especially where turbulent flowsare involved.
REFERENCES
I. SCHLICHTING, HERMANN (J. KESTIN, TRANS.): Boundary Layer Theory. McGraw-
Hill Book Co., Inc., any edition.
2. BIRD, R. BYRON; STEWART, WARREN E.; AND LIOHTFOOT, EDWIN N.: Transport
Phenomena. John Wiley & Sons, Inc., 1960.
3. PRAND'rL, L.: Motion of Fluids with Very Little Viscosity. NACA TM 452, 1928.
4. BLASIUS, H. : The Boundary Layers in Fluids with Little Friction. NACA TM
1256, 1950.
5. HOWARTH, L.: On the Solution of the Laminar Boundary Layer Equations. Proc.
Roy. Soc. (London), Set. A, vol. 164, no. 919, Feb. 18, 1938, pp. 547-579.
6. VON K_.RM._N, TH: On Laminar and Turbulent Friction. NACA TM 1092, 1946.
7. POHLHAUSEN, K.: Approximate Integration of the Differential Equation of the
Limit Surface of Laminar Motion. Zeit. f. Math. Mech., vol. 1, Aug. 1921, pp.
252-268.
8. THWAITE8, B.: Approximate Calculation of the Laminar Boundary Layer. Aero-
naut. Quart., vol. 1, Nov. 1949, pp. 245-280.
9. ROTT, NICHOLAS; AND CRABTREE, L. F.: Simplified Laminar Boundary-Layer
Calculations for Bodies of Revolution and for Yawed Wings. J. Aeron. Sei.,
vol. 19, no. 8, Aug. _,952, pp. 553-565.
10. STEWARTSON, K. : Correlated Incompressible and Compressible Boundary Layers.
Proc. Roy. Soc. (London), Set. A, vol. 200, no. 1060, Dec. 22, 1949, pp. 84-100.
11. COHEN, CLARENC_ B.; AND REStlOTKO, ELI: Similar Solutions for the Compres-
sible Laminar Boundary Layer with Heat Transfer and Pressure Gradient.
NACA TR 1293, 1956.
12. COHZN, CLARZNCZ B.; ANn RESHOTKO, ELI: The Compressible Laminar Boundary
Layer with Heat Transfer and Arbitrary Pressure Gradient. NACA TR 1294,
1956.
13. LUXTON, R. E.; AND YOUNG, A. D.: Generalized Methods for the Calculation of
the Laminar Compressible Boundary-Layer Characteristics with Heat Transfer
and Non-Uniform Pressure Distribution. R&M-3233, Aeronautical Research
Council, Gt. Britain, 1962.
88
INTRODUCTION TO BOUNDARY-LAYER THEORY
14. SMITH, A. M. O.; AND CLUTTER, DARWIN W.: Solution of the Incompressible
Laminar Boundary-Layer Equations. AIAA J., vol. 1, no. 9, Sept. 1963, pp.2062-2071.
15. SMITH, A. M. O.; AND CLUTTER, DARWIN W.: Machine Calculation of Compres-
sible Laminar Boundary Layers. AIAA J., voi. 3, no. 4, Apr. 1965, pp. 639-647.
16. KRAUSE, EaON: Numerical Solution of the Boundary-Layer Equations. AIAA J.,
vol. 5, no. 7, July 1967, pp. 1231-1237.
17. LUDWlEG, HUBERT; AND TILLMANN, W.: Investigations of the Wail-Shearing
Stress in Turbulent Boundary Layers. NACA TM 1285, 1950.
18. MASK_LL, E. C.: Approximate Calculation of the Turbulent Boundary Layer in
Two-Dimensional Incompressible Flow. Rep. AERO 2443, Royal Aircraft
Establishment, Nov. 1951.
19. TRUCKENBRODT, E.: A Method of Quadrature for Calculation of the Laminar
and Turbulent Boundary Layer in Case of Plane and Rotationally Symmetrical
Flow. NACA TM 1379, 1955.
20. RESHOTKO, ELI; AND TUCKER, MAUalCE: Approximate Calculation of the Com-
pressible Turbulent Boundary Layer with Heat Transfer and Arbitrary Pressure
Gradient. NACA TN 4154, 1957.
21. ECKEaT, E. R. G.: Engineering Relations for Friction and Heat Transfer to Sur-
faces in High Velocity Flow. J. Aeron. Sci., vol. 22, no. 8, Aug. 1955, pp. 585-
587.
22. SAaMAN, PHILIP K.; AND CRESCI, ROnERT J.: Compressible Turbulent Boundary
Layer with Pressure Gradient and Heat Transfer. AIAA J., vol. 4, no. 1, Jan.
1966, pp. 19-25.
23. McNALLY, WILLIAM l).: FORTRAN Program for Calculating Compressible
Laminar and Turbulent Boundary Layers in Arbitrary Pressure Gradients.
NASA TN 1)-5681, 1970.
24. HERRING, H. JAMES; AND MELLOR, G. L.: A Method of Calculating Compressible
Turbulent Boundary Layers. NASA CR-1144, 1968.
25. CEnECl, T.; SMITH, A. M. O.; AND MOSINSKIS, G.: Solution of the Incompressible
Turbulent Boundary-Layer Equations with Heat Transfer. J. Heat Transfer,
vol. 92, no. 1, Feb. 1970, pp. 133-143.
26. SMITH, A. M. O.; AND CEnEcI, T. : Numerical Solution of the Turbulent-Boundary-
Layer Equations. Rep. I)AC-33735, Douglas Aircraft Co. (AD-656430), May1967.
27. CEnECI, T.; AND SMITH, A. M. O.: A Finite-Ditference Method for Calculating
Compressible Laminar and Turbulent Boundary Layers. J. Basic Eng., vol. 92,
no. 3, Sept. 1970, pp. 523-535.
28. BRADSHAW, P.; FERRISS, D. H.; AND ATWELL, N. P.: Calculation of Boundary-
Layer Development Using the Turbulent Energy Equation. J. Fluid Mech.,
vol. 28, pt. 3, May 26, 1967, pp. 593-616.
29. BRADSHAW, P.: Calculation of Boundary-Layer Development Using the Turbu-
lent Energy Equation. IX: Summary. Rep. NPL-Aero-1287, National Physical
Lab., Jan. 30, 1969.
30. PATANKAR, S. V.; AND SPALDING, D. B.: A Finite-Difference Procedure for Solving
the Equations of the Two-Dimensional Boundary Layer. Int. J. Heat Mass
Transfer, vol. 10, no. 10, Oct. 1967, pp. 1389-1411.
31. PATANKAR, S. V.; AND SPALDING, D. B.: Heat and Mass Transfer in Boundary
Layers. C.R.C. Press, 1967.
32. COLES, D. E.; AND HIRST, E. A., EDS.: Proceedings, Computation of Turbulent
Boundary Layers--1968, AFOSR-IFP-Stanford Conference. Stanford Univ.
Press, 1969.
89
TURBINE DESIGN AND APPLICATION
33. BERTRAM, 1V_ITCHEL H., ED.: Compressible Turbulent Boundary Layers. NASA
SP-216, 1969.
9O
INTRODUCTION TO BOUNDARY-LAYER TtIEORY
SYMBOLS
A, turbulent flow mixing coefficient, (N) (sec)/m2; lbm/(ft) (see)a constant in eq. (6-44)
(width of flat plate, m; ftb _constant in eq. (6-44)
CI skin-friction coefficient for a flat platec constant in eq. (6-44)
c_ specific heat at constant pressure, J/(kg)(K); Btu/(lbm)(°R)D total drag on flat plate, N; lbf
d constant in eq. (6-44)
f Blasius dimensionless stream function defined by eq. (6-63)f general body force vector, N/kg; lbf/lbm
f_ component of body force f in x-direction, N/kg; lbf/lbm
f_ component of body force f in y-direction, N/kg; lbf/lbm
f, component of body force f in z-direction, N/kg; lbf/lbm
g conversion constant, 1; 32.17 (lbm) (ft)/(lbf) (seC)H form factor, defined by eq. (6-56)i unit vector in the x-direction
J conversion constant, 1 ; 778 (ft) (lbf)/Btu
j unit vector in the y-direction
k thermal conductivity, W/(m) (K) ; Btu/(sec) (ft) (°R)k unit vector in the z-direction
L characteristic length (e.g., the blade chord), m ; ft
fPrandtl mixing length, m; ftl _length of flat plate, m; ft
M, Mach number external to the boundary layer
n exponent on the turbulent velocity profile, eq. (6-47)
P dimensionless pressure, defined by eq. (6-13e)
p static pressure, N/mS; lbf/ft 2
R gas constant, J/(kg) (K); (ft) (lbf)/(lbm) (°R)
Re Reynolds number based on L and U0, as defined by eq. (6-13f)Re_ Reynolds number based on l, as defined in eq. (6-75)
Rex Reynolds number based on x, as defined by eq. (6-67)
r radius of curvature of blade surface, m; ft
S constant in eq. (6-25), K; °R
T absolute static temperature, K; °R
Tt absolute total temperature, K; °R
To reference temperature used in eq. (6-25), K; °Rt time, sec
U dimensionless velocity in x-direction, defined by eq. (6-13c)
U0 free-stream velocity upstream of blade, m/see; ft/sec
u component of general velocity vector u in the x-direction, m/sec;ft/sec
91
TURBINE DESIGN AND APPLICATION
u general velocity vector, m/see; ft/sec
u, free-stream velocity at the outer edge of the boundary layer,
m/sec; ft/secV dimensionless velocity in y-direction, defined by eq. (6-13d)
v component of general velocity vector u in the y-direction, m/sec;
ft/sec
w component of general velocity vector u in the z-direction, m/see;
ft/secX dimensionless x-coordinate, defined by eq. (6-13a)
ix-coordinate, m; ftx _coordinate parallel to boundary surface, m; ftY dimensionless y-coordinate, defined by eq. (6-13b)
[y-coordinate, m; ftY _coordinate perpendicular to boundary surface, m; ft
z z-coordinate, m; ft
displacement thickness, m; ft
_f_u boundary-layer thickness, m; ft
[eddy viscosity defined by eq. (6-80), m2/sec; ft2/sec_,a dimensionless quantity much less than 1
Blasius transformed y-coordinate defined by eq. (6--64)0 momentum thickness, m; ft
dimensionless shape parameter defined by eq. (6-45)
dynamic viscosity, (N) (sec)/mS; lbm/(ft) (see)m dynamic viscosity at reference temperature To, (N)(sec)/m2;
Ibm/(ft) (see)v kinematic viscosity, m2/sec; ft2/sec
p density, kg/m 3; lbm/ft _p, free-stream density external to the boundary layer, kg/m*;
lbm/ft 3rl laminar shear stress, N/m2; lbf/ft s
rt turbulent shear stress, N/mS; lbf/ft 2• w shear stress at the wall, N/mS; lbf/ft 2
function defined by eq. (6-34)
_b Blasius stream function, m2/sec; ft2/sec
constant in eq. (6-26)
Superscripts:
time average
fluctuating component
92
CHAPTER7
Boundary-LayerLosses
By HermanW.Prust,Jr.
The primary cause of losses in a turbine is the boundary layer thatbuilds up on the blade and end-wall surfaces. In particular, these losses
are the friction loss resulting from the flow of the viscous fluid over the
surfaces, the pressure-drag loss resulting from the flow of fluid past the
blade trailing edge, and the loss downstream of the blades resulting from
the mixing of the low-velocity boundary-layer fluid with the high-velocity
free-stream fluid. Chapter 6 presented an introduction to boundary-layer
theory, by means of which the surface boundary-layer buildup can be
analytically described. This chapter covers analytical and experimental
methods for determining the friction, trailing-edge, and mixing losses
associated with the boundary layer. The theory presented herein refers
primarily to two-dimensional blade-section boundary layers. Methods for
obtaining three-dimensional blade plus end-wall losses from the two-dimensional results are also discussed.
A fundamental objective in blade-row design is to minimize the energy
loss resulting from the flow of fluid through the blade row. Therefore, the
final expressions for loss developed in this chapter are in terms of kinetic-
energy loss coefficients. These coefficients express the loss in fluid kinetic
energy as a fractional part of the ideal kinetic energy of the actual flow
through the blade row. Efficiency based on kinetic energy can be obtained
by subtracting these coefficients from unity, and this is consistent with
the blade-row efficiency definition used in chapter 2.
Before proceeding with the discussion of boundary-layer parametersand loss coefficients, the blade-row station locations and associated
pressure and velocity distributions will be introduced with the aid of figure7-1. These pressure and velocity distributions and the associated dis-
93
TURBINE DESIGN AND APPLICATION
cussion refer to an attached boundary layer only. A separated boundary
layer, with its associated reversal of flow at the surface, is thicker, yields a
higher loss, and cannot be analyzed in the same manner, if at all.
Figure 7-1(a) indicates the four station locations that will be referred
•_---- S------_
(a)
Station
0
r la
Station 0
-- Total pressure
.... Static pressure
--- Velocity
Vfs.la /
Station la
Station 1
......tStation 2
_b_
(a) Station locations.
(b) Pressure and velocity distributions.
FIGUaE 7-1.--Station locations and associated pressure and velocity distributions.
94
BOUNDARY-LAYER LOSSES
to in this chapter. Station 0 represents the inlet to the blade row. At this
station, a uniform total pressure po' is assumed, as indicated in figure
7-1(b). Station la is just within the trailing edge of the blade. The
boundary layers developed on the blade surfaces result in velocity and
pressure profiles as shown in figure 7-1 (b). Velocity varies from the free-
stream value Vj,.1, to zero at the blade surfaces. There is, of course, no
flow through the region of the solid trailing edge. Total pressure varies
from the free-stream value p_,,1, = p0' to the static pressure pl_ at the blade
surfaces. This static pressurc is assumed constant across station la, as is
the flow angle al,. At station la, only the surface friction loss has occurred.
Station 1 is just beyond the blade trailing edge, where the boundary-
layer fluid has filled the void, but where little mixing with the free stream
has occurred. This is indicated in figure l(b) by the station-1 profiles
showing flow throughout the entire wake region. Here too, static pressure
and flow angle are assumed constant across the station. Between stations
la and 1, the trailing-edge loss occurs. Station 2 is located at a distance
sufficiently downstream of the blade rot. that completc mixing, with the
associated mixing loss, has taken place. The velocity and total-pressure
profiles are again uniform.
In order to simplify analysis and discussion, a number of variableshave been assumed constant across the various stations. Uniformity of
inlet conditions is a universal convenience that usually can be approached
in component tests but seldom exists in actual applications. Experiments
have shown that static pressure and flow angle do vary somewhat acrossboth free stream and boundary layer at stations la and 1. In some in-
stances, which will be later identified, this variation can be accounted for.
Although some downstream mixing of the flow does take place, a com-
pletely uniform downstream state is merely a hypothetical convenience.
BOUNDARY-LAYER PARAMETERS
When a real fluid riot's over a surface, a loss results due to both frictionbetween the fluid and the surface and friction between the layers of fluid
in the region adjacent to the surface. As shown by figure 7-2, the fluid
velocity in the boundary-layer region varies from zero velocity on the
surface to free-stream velocity Vf, at the full boundary-layer height
_I_u. To describe the losses in flow, momentum, and energy resulting from
the presence of the boundary layer, certain parameters are used. Some of
these (displacement thickness, momentum thickness, form factor) were
introduced in the last chapter and will be reviewed here; in addition, others
specifically used for obtaining the desired kinetic-cnergy coefficients willbe introduced and defined.
95
TURBINE DESIGN AND APPLICATION
Free-stream velocity,
Vfs iJ
Velocity, V _;/
_/Surface
," Full boundary
layer height. 6full
FIGURE 7-2.--Typical boundary-layer velocity profile.
The displacement thickness 6, which is indicative of the loss in mass
flow, is defined by
6full LS/ull_(pV)I. = (pV)], dY-- (pV) dY (7-1)
v 0
where
6
_f=uV
P
Y
():.
displacement thickness, m; ft
boundary-layer thickness, m; ft
fluid velocity, m/sec; ft/sec
fluid density, kg/m3; lb/ft _
distance in direction normal to boundary layer, m; ft
free-stream (ideal) conditions
Equation (7-1) states that the loss in mass flow of the fluid in the bound-
ary layer is equal to the ideal flow which would pass through a length
(or an area) equal to the displacement thickness. Solving for _ yields
f:=" f/=" pVdY-- (7-2)=-o vo (_-)/ dY
The momentum thickness 0, which is indicative of the momentum
loss, is defined by
fsj., dy_ fo_:=.O(pV2):" =-o (pVV:,) pV 2dY (7-3)
where 0 is the momentum thickness, in m or ft. Equation (7-3) states that
the loss in momentum of the fluid in the boundary layer is equal to the
96
BOUNDARY-LAYER LOSSES
ideal momentum of the ideal flow which would pass through a length (or
an area) equal to the momentum thickness. Solving for 0 yields
0 fs/=zt oV dy_ff/"'t pV'- dY (7-4)=-o (pv):. (pv')/---_.
The lossin kineticenergy can be similarlyexpressed in terms of an
energy thicknessdefinedby
f':"" 1["""- (pV 3) dY (7-5)1 _b(pVa)/.= _ (pVV_,) dY--_-o2 "o
where _b is the energy thickness, in m or ft. Equation (7-5) states that
the loss in kinetic energy of the fluid in the boundary layer is equal to theideal kinetic energy of the ideal flow which would pass through a length
(or an area) equal to the energy thickness. Solving for ¢ yields
f,:.z, pV ,/.z_='o] (P-_/_ dY- fo PVa¢
(pV3) :' dY (7-6)
Ratios of the aforementioned thickness terms are also used as basic
boundary-layer parameters. The form factor H is defined as
U = - (7-7)0
Substituting equations (7-2) and (7-4) into equation (7-7) and defining a
dimensionless distance y as
Yy = -- (7-8)
yields
1 1
d,-/0(pW)f$
H - (7-9)
fo I pV . [i pV 2(_-): aY-Jo (p-_): dy
An energy factor E is defined as
E= -_ (7-10)O
Substituting equations (7-6), (7-4), and (7-8) into equation (7-10)
yields
97
TURBINE DESIGN AND APPLICATION
_o_ pV , fl pV _ dy(_-)i. aY-Jo (pv3),_
E=
1 p V _o 1 dv(pV_)f.
(7-11)
Velocity profiles for turbulent flow are often represented by a powerprofile of the type
V-y_ (7-12)vj.
where the value of the exponent n is most often between 0.1 and 0.25.
Note that this power profile is here expressed as YL while the same
profile in chapter 6 (eq. (6-47)) is expressed as y_/n. The exponent ex-
pressed as 1In is consistent with general boundary-layer theory usage.
The exponent expressed as n, however, is consistent with reference 1,wherein the equations that follow are derived. Therefore, the specific
numerical value to be used for n will depend on the form being used for
the exponent.
With this velocity profile, equations (7-9) and ( 7-11 ) can be integratedin series form, and the form and energy factors for turbulent flow can be
expressed in terms of the exponent n and the free-stream eritical velocityratio V/Vcr. The resulting equations derived in reference 1 are
1 3At, 5A_,
H- (7-13)
and
E=
1 AI. A_.
(n._F1)(2n+l)_-(3n+l)(4n+l) T (5n_F1)(6n+l) t-'..
1 A I,2 (n+l)_3n+l)+(3n+l)(5n+l)+
]
(5n+l) (7n+l) J
where
1 A s,
(n+l) (2n+l) + (3n+l) (4n+l)+
AL(Sn+ 1) (6n+l) +""
AI. - _- 1
(7-14)
(7-15)
98
BOUNDARY-LAYER LOSSES
and _, is the ratio of specific heat at constant pressure to specific heat at
constant vMume, and Vcr is the fluid velocity, in m/sec or ft/see, at the
critical (Mach 1) flow condition. For incompressible flow, where V/V,
approaches zero, equations (7-13) and (7-14) reduce to
andH,,c=2n+ l (7-16)
2(2n-F1)
E,,¢- 3n+-----_ (7-17)
Values of the form and energy factors for turbulent compressible flow
are shown in figure 7-3 for V/V, varying from 0 to 1.4 and n varyingfrom 0 to 1.5. It can be seen that the form factor varies much more than
does the energy factor. For any constant exponent n, the energy factoris almost independent of V/V,.
The boundary-layer parameters just presented are general and can
refer to a boundary layer on any type of body. They are directly useful incertain aerodynamic work. For instance, the drag of a body can be ob-
tained directly from the momentum thickness. In turbine work, however,
where the flow is confined to the physical boundaries of the blade row, it is
simpler and more meaningful to express the losses as a fractional part of
the ideal quantities that could pass through the blade row. The thickness
parameters so expressed are herein termed "dimensionless thickness
parameters" and are defined on the basis of zero trailing-edge thickness.
W
t.
Free-st reamcritical-
velocity
ratio,
(VNcr)fs
r0///- 0.6
-- 1.0 PowernZ 2 /////r// , 1.4 7 ,-0 used in
.... .' ' velocity
_k _ ,,-O.I0 " equation
1.81-\\\ \I "__ ro.5o ,-o75 _l.oo/ -_-_ _ , " r125
.... , /. /-iSO].4| I r--_T- i , I1 2 3 4 5 6 7
Formfactor,H
FIGURE 7-3.--Effect of compressibility on variation of energy factor with form factor.
(Data from ref. I.)
99
TURBINE DESIGN AND APPLICATION
t 5
°1a
FIGURE 7-4.--Nomenclature for trailing-edge region.
These dimensionless thickness parameters must represent the sum of the
suction- and pressure-surface thicknesses.
With the assumption that flow conditions in all channels are the same,
the dimensionless thickness parameters are obtained by dividing the
losses in flow, momentum, and energy for a single blade-row channel by
the corresponding ideal quantities that could pass through one blade-row channel. The total losses for one channel, as indicated in figure 7-4,
are composed of the suction-surface loss plus the pressure-surface loss, or
8tot = 8.+8p (7-19)
4J_ot--_b.+ _, (7-20)
where the subscripts tot, s, and p denote total value, suction-surface value,
and pressure-surface value, respectively. Thus, in terms of the previously
defined boundary-layer thicknesses, the dimensionless boundary-layer
thicknesses are expressed as
$,o,(pV)f. _,°,$*= - (7-21)
s cos a(pV)f, s cos a
#,o,(pV2)/. 0,o,#* = - (7-22)
s cos a(pV2)/, s cos a
_to__b*.... (7-23)
8 COS
100
where
6"
8"
8
BOUNDARY-LAYER LOSSES
dimensionless displacement thicknessdimensionless momentum thickness
dimensionless energy thickness
blade spacing, m; ft
fluid flow angle from axial direction, deg
Equations (7-21), (7-22), and (7-23) express the losses in flow, momen-
tum, and energy, respectively, as fractions of their respective ideal quanti-ties for the blade row if the trailing-edge thickness is assumed to be zero.
These equations can be subscripted to apply at either station la, withinthe trailing edge, or station 1, beyond the trailing edge.
BLADE-ROW LOSS COEFFICIENTS
As mentioned previously, the losses are to be expressed in terms of
kinetic-energy loss coefficients. In this section, methods for evaluating thefriction, trailing-edge, and mixing losses and expressing them in terms of
the kinetic-energy loss coefficients will be presented.
Surface-Friction Losses
The kinetic-energy loss coefficient _1,, defined as the loss in kinetic
energy as a fraction of the ideal kinetic energy of the blade-row actual
flow, can be expressed in terms of the boundary-layer dimensionlessthicknesses as
*_l.s cos a_. (pV 3)i,,1a
e_"-- (s cos al.--_*as cos ale--t) (pVa)1.a, (7-24)
where t is the blade-row trailing-edge thickness, in m or ft. (Refer to fig.
7-4 for the nomenclature in the region of the trailing edge of the blade.)
Since this coefficient is referenced to station la, just within the blade-row
trailing edge, it represents only the surface-friction loss. If a trailing-edgedimensionless thickness is expressed as
tt* = _ (7-25)
S COS O_la
equation (7-24) reduces to
$_°= 1 --$*.-- t* (7-26)
In order to evaluate the loss coefficient e_. from equation (7-26), it is
necessary to know the values of the dimensionless energy thickness _b_*_
101
TURBINE DESIGN AND APPLICATION
and the dimensionless displacement thickness 51"- These can be evaluatedeither experimentally or analytically, as will be discussed herein.
Experimental determination.--In determining experimental loss values,
it is impractical to measure the density and velocity directly. Instead,
pressure data are taken, and the density and velocity are related to
pressure functions. The pressure data required for computing the friction
loss consist of (see fig. 7-1) the upstream total pressure p0', the blade-exit
static pressure pl., and the total-pressure loss survey data po'-p'_o for one
blade space Since the dimensionless boundary-layer thicknesses expressthe losses of the blade row as a fractional part of the ideal quantities which
could pass through the blade row, the dimensionless displacement thick-
ness can be expressed in terms of the flow across one blade pitch as
m
s cos al,(pV)i,a,--t*s cos al,(pV)/,a_-eos al_ fo (pV)I. du
_1"= (7-27)s cos al,(pV)I,._,
where u is the distance in the tangential direction, in m or ft. Equation(7-27) simplifies to
C)61"=l-t,-[ .v d (7-28)_0 la
In a similar manner, the dimensionless momentum and energy thick-nesses can be expressed as
81"=
1 _ 1
(PV2)I,,I=
i 1 V (:) (7-29)
and
1 U 1
*_/la
(pI.V].) 1=
V _ pV=fo [1-- (V-_I,)I,] (_),, d (u) (7-30)
Assuming that the total temperature T' and the static pressure p_ in
102
BOUNDARY-LAYER LOSSES
the boundary layer are the same as in the free stream, the density ratio
(P/Pl.) 1_ can be related to the pressure ratio p_./po t as follows: From the
isentropic relation,
(_),. = \_]1. (7-31)
p "p ,,1Iv
Since ' ' ' ' ' 'p1..1.=po and pidpl..1,,--pl°/po (from the ideal gas law, with
T'I.= T'I.._.= To'), division of equation (7-31) by equation (7-32) yields
The velocity ratio (V/VI.)_. can be related to the pressure ratios
(pip') _. and p_Jpo' as follows: From the total-temperature definition and
isentropic relation, equations (1-51) and (1-52) of chapter 1, we canwrite
V 2 T (p _(,-1)/,2gJcvT , l-v-- 1-\_/ (7-34)
conversion constant, 1; 32.17 (lbm) (ft)/(lbf) (sec 2)conversion constant, 1; 778 (ft) (lb)/Btu
specific heat at constant pressure, J/(kg) (K); Btu/(lb) (°R)
where
gJ
Cp
Subscripting equation (7-34) once for station la and again for free-stream
values at la, dividing the first of these equations by the second, and
recalling that p'] .,_. = po' and T_ .,_°= T'I,, yields
1- --;-\PxJV
,,-,,,,
With the density and velocity ratios expressed in terms of the measured
pressures by equations (7-33) and (7-35), it is now posaible to integrate
equations (7-28), (7-29), and (7-30) and evaluate the dimensionless
boundary-layer thicknesses. Then, the kinetic-energy loss coefficient _1.
can be computed from equation (7-26).
The kinetic-energy loss coefficient thus determined is a two-dimensional
103
TURBINE DESIGN AND APPLICATION
coefficient; that is, it is based on data either from a two-dimensional
cascade or from a constant radius of an annular cascade. (The annular
cascade can be, and often is, the full stator or rotor from a turbine.) In
order to obtain a three-dimensional loss coefficient for a blade row, data
are taken at a number of radii sufficient to adequately cover the annulus,
and the two-dimensional dimensionless boundary-layer thicknesses are
calculated as shown previously for each radius. Three-dimensional bound-
ary-layer thicknesses are then obtained by radial integration from hub
to tip:
L " 5* ( pV ) ]. ,I. cos al. r drh
5* --la,3D --
" (pV) s.,l.cosa_ r drh
(7-36)
' O_(PV2)/.,t_ cos alo r dr
0_,,D= (7-37)
' (pV2) [.,,_ cos at. r dr
' _b*a(PV') /.,1. cos alo r dr
_b*,,D-- (7-38)
., (PV3) s.,l. cos alo r drk
In terms of the measured pressures, these integrals are expressed as
"'c.(p,o),/,Pl (p_o_(v-_)Iv]_/sL,, L -\_/ J cos ala r dr
51",,D-
S,:'<.,=>,,.[,_c.-v'-""l'"\_I j cosal.rdr
(7-39)
104
BOUNDARY-LAYER LOSSES
, e*(p,_) '/* 1--\_ot ] jcosai.rdr
", [1_ fpl.y*-"/_]L, (PI')I/" L \p01/ .] cosal.rdr
(7-40)
L, _bl*'(P':)l/" L \_/ 2 cosaiordr
_bi.,aD-- (7--41)
L:' (P'<')'7"t[ 1-fpi"_¢'r-')/'r1311\_/ j cos a,: r dr
The three-dimensional kinetic-energy loss coefficient is then obtained
in a manner similar to equation (7-26) :
_la,$D
_i.,3D _ 1 -- _*o,ID-- t,.* (7--42)
where t,* is the trailing-edge dimensionless thickness at the mean radius
and is used to represent the average value for the blade row.Analytical determination.--The kinetic-energy loss coefficient ll. can
also be evaluated with the use of analytically determined boundary-layer
thickness parameters. While not as reliable as experimental values, ana-
lytical values are much less costly and time consuming to obtain. Ana-
lytical methods for calculating the basic boundary-layer parameters arediscussed and referenced in chapter 6. The boundary-layer solutions are
not simple, and the better methods require computer solution. Boundary-
layer computer programs currently in use at the NASA Lewis Research
Center include one (ref. 2) based on an integral method solution andanother based on the finite difference method of reference 3.
An equation used in the study of reference 4 to compute turbulent
boundary-layer momentum thickness was
01a
0.231
<,+-,->:_.c.,.,.
×Io!
l-f,>v ]o.,--- -- (1 -As.) °'_7 dxLkp v=J/. \V<,ls. J ,,pvIs.
lO O,Sli(ln+i)
0.7886
(7-43)
105
TURBINE DESIGN AND APPLICATION
where
AI° parameter defined by equation (7-15)
x distance along blade surface from forward stagnation point, m; ft
l blade-surface distance from forward to rear stagnation point,m; ft
viscosity, (N) (sec)/m2; lb/(ft) (sec)
The development of this equation is presented in reference 4. It is as-
sumed that the boundary layer has a power-law velocity profile. In
reference 4, the exponent n was obtained from the referenced equation
1TM1 2.6 x (7-44)n LkP- lya _I
Equation (7-43) must be evaluated for both the suction and pressure
surfaces of the blade. The free-stream velocities and densities required for
equations (7-43) and (7-44) are those free-stream values adjacent to the
blade-surface boundary layers. These can be obtained by any of thechannel flow analysis techniques discussed in chapter 5.
Values of the form factor H as required in equation (7-43) and of theform factor H,_ and energy factor EI_ at station la for each surface can be
obtained from equations (7-13) and (7-14). With 0_a, Hla, and E,_ known
for both the suction and pressure surfaces, the various boundary-layerthickness parameters and the kinetic-energy loss coefficient $_a can be
evaluated from the equations presented earlier in this chapter. For the
turbine stator blade studied in reference 4, the analytical values, as
calculated from equation (7-43), of the boundary-layer momentum
thickness for the blade and for the two surfaces individually were reason-
ably close to the experimental values. In general, however, results ob-tained from equation (7-43) will not be as accurate as those obtainedfrom the computer programs of references 2 and 3.
Three-dimensional boundary-layer parameters could be calculated
directly from equations (7-36) to (7-38). The two-dimensional thickness
parameters would have to be analytically determined at a number of radii
sufficient to establish the variation over the blade length and would also
have to be determined, somehow, over the end-wall surfaces. Such a
procedure would require considerable effort, so the simplified method ofreference 5 for predicting three-dimensional losses from two-dimensional
mean-section losses is commonly used. Results obtained by this methodhave shown good agreement with experimental results.
In the method of reference 5, the following assumptions are made:
(1) The average momentum loss for the blade surface can be represented
by the dimensionless momentum thickness at the blade mean section;
106
Approximatearea_one endwall
Aw - sccos%1-7
BOUNDARY-LAYER LOSSES
Approximateareaofone
FIGURE7-5.--Schematic diagram of equivalent two-dimensional blade used to calcu-late the effect of end-wall area on blade loss.
(2) the momentum loss per unit area on the inner and outer end walls is
the same as the average momentum loss per unit area on the blade surface;
and (3) the blade configuration can be satisfactorily approximated by an
equivalent two-dimensional blade, as shown in figure 7-5, having a con-stant cross section, spacing, and stagger angle equal to those at the mean
section of the given blade. The surface area of one equivalent blade(see fig. 7-5) is
A b= 2ch (7-45)where
Ab total surface area (sum of suction-surface and pressure-surfaceareas) of one blade, mS; ft 2
c blade chord, m; ft
h blade height, m; ft
The inner and outer end-wall area for one passage is
where
A_
Aw = 2sc cos a, (7-46)
total surface area of passage end walls (sum of inner and outer
end-wall areas), m2; ft 2
a, blade stagger angle, deg
Now, taking the average momentum loss 0"1_,,_over the blade radial length
107
TURBINE DESIGN AND APPLICATION
and modifying it to include the end-wall losses yields
\{Ab+A_--O*.Ab/ ( scosa,\01*m) = O*,o,m - +--T--) (7-47)
The three-dimensional energy and displacement thickness parametersare then calculated as
*= EIo,.,OI_,3D (7-48)
and
la,3D- Hla,,.Ola,aD (7-49)
Mean-section values are used for the energy and form factors. Althoughthe energy and form factors were originally defined in terms of individual
boundary-layer thicknesses, it is indicated in reference 4 that they can besatisfactorily used as is done in equations (7-48) and (7-49). The three-
dimensional kinetic-energy loss coefficient is then obtained from equa-tion (7-42).
Trailing-Edge Loss
The kinetic-energy toss coefficient 6e that represents the loss associated
with flow past the blade trailing edge can be determined either experi-mentally or analytically.
Experimental deterrnination.--Experimental values of blade trailing-edge loss coefficient _t, are obtained from differences between experimental
two-dimensional loss coefficients _t, which include both surface-friction
loss and trailing-edge loss, and loss coefficients _1_,which include only theblade surface-friction loss. Thus,
= (7-50)
Loss coefficients _o, which include only surface-friction loss, are ob-
tained as described previously. Loss coefficients *_, which include both
surface-friction loss and trailing-edge loss, are determined in exactly the
same manner except that the total-pressure loss and static pressure aremeasured at different locations. The surface-friction loss coefficients were
based on data obtained just within the blade trailing edge at station la,
where the trailing-edge loss has not yet occurred. To determine the loss
coefficients which include both surface-friction loss and trailing-edge loss,
the measurements must be made at a location just downstream of the
blade row, corresponding to station 1 in figure 7-1, where the trailing-edge
loss, but little mixing, has occurred.
Analytical determination.--In reference 6, experimental drag coefficients
are presented for a large number of surface discontinuities. Included in
the reference are experimental data for sheet-metal joints of different
108
BOUNDARY-LAYER LOSSES
{pV21fsl2g _i
v/////////,_////////////////////z
FIGURE 7-6.--Schematic diagram of body in boundary layer.
geometry, bolt and rivet heads of different geometry, and airfoil trailing
edges. It is indicated that the pressure-drag loss due to the discontinuity
behaves similarly regardless of the flow direction over the discontinuity.
Therefore, the loss due to flow past a trailing edge will be treated ana-
lytically as if the loss were due to a body placed in the path of a boundary
layer.As indicated in reference 6, the drag of a small body of height h, equal
to or less than the full boundary-layer height 5].m placed in a turbulent
boundary layer, as shown in figure 7-6, corresponds approximately to the
effective dynamic pressure of the part of the boundary layer equal to the
height of the body. Thus,
where
D
h
CD
D = q_//hCD (7-51)
drag on body, N/m; lb/ft
height of body, m ; ft
drag coefficient
and the effective dynamic pressure q,z: is expressed as
1 foh pV 2dY (7-52)qe:/=h 29
Drag is related to momentum thickness as
D= O(pV_):" (7-53)g
Therefore, a dimensionless momentum thickness 0"_ representing the
trailing-edge loss is obtained by combining equations (7-51) and (7-53)
with a lbroperly subscripted form of equation (7-22) :
109
TURBINE DESIGN AND APPLICATION
O,te __ qey/hCD (7-54)
(pV2):.8 COS Oq --
g
The flow angle _1 is related to the angle al_ as discussed in chapter 4 (eqs.(4-26) and (4-27)).
Before equation (7-54) can be evaluated, the effective dynamic pres-
sure must be determined. The ratio of the effective dynamic pressure tothe free-stream dynamic pressure is equal to
q:. h dY (7-55)
For turbulent flow, the variation of velocity in the boundary layer can be
expressed with the use of the simple power profile presented previously.Combining equations (7-8) and (7-12) yields
V:. \_/,m/ (7-56)
Assuming that the total temperature and static pressure in the boundary
layer and frec stream are the same and using the ideal gas law and equa-tion (1-64) of chapter 1 gives
To' "y+ 1. Vcr.:.
To' 7+1 I7_
(7-57)
Substituting equations (7-56) and (7-57) in equation (7-55) and usingthe parameter A:. defined by equation (7-15) yields
h
qe:] 1 (I_A:.) 1 ( Y _'"qf.-h \_t,,t,./ dY (7-58)
(YY Il-A:. \6:../ J
o
Performing a binomial expansion and integrating then gives
r/h v- 1 : h A s.q'f:-(1-A:.)q]. [_6_z,) 2-_+_5-_,m) (4n+l)
+( h_']'" A_-. )+...] (7-59)\6:,,u/ (6n+l
110
BOUNDARY-LAYER LOSSES
Substituting equation (7-59) in equation (7-54) and using trailing-edgethickness t in place of body height h finally yields
. 1 A,.tCD (l--A/,) +Ot*-2s cos ax L\_f, zt/ _ _t (4n+l)
+ (6n+l_+... (7-60)
The boundary-layer thickness _i,u to be used in equation (7-60) should
be the sum of the suction- and pressure-surface values.
Equation (7-60) is for compressible flow. In many cases, at least when
n is not well known, the following simplified equation, which assumes
incompressible flow and n=l/_ (commonly used for turbulent flow),
is adequate:
e*=o.3752/t_ re. (7-61)_futl 8 COS Oil
The information in reference 6 indicates that the drag coefficient CD can
be set equal to 0.16 for a rounded trailing edge and 0.22 for a square
trailing edge. The corresponding values reported in reference 7 and con-
verted to the same basis as equation (7-61) arc 0.14 for a rounded trailing
edge and 0.22 for a square trailing edge. Frequently, _,o, instead of _iuu
will be available. In such a case, for incompressible flow,
/f/,,u = (5,ot(_-1-1) (7-62)
and for compressible flow,
_tot
_fuu (7-63)
[ 1 A f, A_,1--(l-A/,) _n-_+3-_+5-_+- .-)
Equations (7-60) and (7-61) give the fractional loss in blade-row
momentum due to the blade trailing edge. To find the corresponding
kinetic-energy loss coefficient, it is necessary to find the fractional losses
in flow and kinetic energy. As a simple approximation, the form and energy
factors, evaluated from equations (7-13) and (7-14) for compressible
flow and from equations (7-16) and (7-17) for incompressible flow, areused to obtain
_*e= H O*, (7-64)
and* *
_b,, = EO,, (7-65)
111
TURBINE DESIGN AND APPLICATION
At station 1, which is just downstream of the blade trailing edge, fluid has
flowed into the area behind the trailing edge and there is no longer a void
due to trailing-edge blockage• Therefore, a kinetic-energy loss coefficientis obtained as
¢*,e,.= 1_8. _ (7--66)
This loss coefficient expresses the loss in kinetic energy as a fraction of the
ideal kinetic energy of the flow that would exist if the trailing-edge loss
were the only loss. The trailing-edge kinetic-energy loss coefficient for
incompressible flow is plotted against trailing-edge thickness in figure 7-7
for several values of the ratio of trailing-edge thickness to boundary-layer
thickness. This figure is based on the momentum loss as expressed by
equation (7-61). The flow loss associated with blade-surface friction is
not included in equation (7-66). Therefore, this trailing-edge kinetic-
energy loss coefficient is approximately, but not rigorously, additive with
the surface-friction loss coefficient. Expression of the combined frictionand trailing-edge loss in terms of a kinetic-energy loss coefficient is dis-cussed in the next section.
o
c"a
'Eb--
0,, /ii//
•010-- J_ . 1
•_5_
0 .05 .10 .15
Dimensionlesstrailing-edgethickness,t*
Ratio of
trai li ng-edge
thickness to
boundary-layer
height,
t/_ull
FIGURE 7-7.--Effect of trailing-edge blockage on kinetic-energy loss coefficient. Form
factor H = 1.3; energy factor E = 1.8; drag coefficient CD =0.16.
112
BOUNDARY-LAYER LOSSES
Combined Friction and Trailing-Edge Loss
As stated in the discussion of the trailing-edge loss, the boundary-layer
thickness parameters and a kinetic-energy loss coefficient expressing the
combined friction and trailing-edge loss can be obtained by making the
experimental measurements at a location corresponding to station 1,
which is just downstream of the trailing edge. In this way, we obtain
experimental values of $1", 01", and _bl* from appropriately subscripted
versions of equations (7-28), (7-29), and (7-30). The value of 61 is thenobtained as
$1- (7-67)1 - $1"
Analytically, the boundary-layer thickness parameters at station 1 are
obtained by adding the surface-friction loss to the trailing-edge loss.
Before the friction and trailing-edge boundary-layer thickness parameters
can be added, they must be expressed on the basis of the same ideal flow.
The friction-loss dimensionless thicknesses at station la (_1", 01*, _bl*) are
expressed in terms of an ideal flow without trailing-edge blockage. How-
ever, there is a trailing-edge blockage at station la, where the ideal flow
with blockage must be comparable to the ideal flow at station 1, where
there is no blockage. Therefore, the friction-loss boundary-layer thickness
parameters are adjusted to account for the true (with blockage) idealflow as follows:
$, _$. ( s cos,l, _ (7-68)1,I- lo \S COS o la--g/
and
0" -0" ( s c°s a_° (7-69)
/ \, , ( s_cos al___) (7-70)
_1.1 = ¢/1. \s cos al.-t�
where the subscript f refers to the loss due to surface friction. Adding the
friction and trailing-edge loss parameters then yields the combined loss
parameters at station 1:
* * (7-71)$1" = $1,i+$ t.
01"= 01,f_O,. (7-72)and
_1* * *= _bl,y_-_b,, (7-73)
And the value of _1 is then obtained from equation (7-67).
113
TURBINE DESIGN AND APPLICATION
After-Mix Loss
The after-mix loss is the total loss that includes the surface-friction loss,the trailing-edge loss, and the mixing loss. The after-mix loss coefficient
$2is determined as described in this section, and the mixing loss, if desired,is obtained by subtracting the previously determined St from $2.
To determine the after-mix loss experimentally would require that the
pressure measurements be made downstream of the blading where com-
plete mixing has occurred. This is impractical for several reasons: (1)
The length for complete mixing, while quite long, is unknown; (2) theafter-mix loss would have to be corrected for side-wall friction for the
mixing length, thus leading to possible error; and (3) after the flow had
mixed, values of after-mix po'-p2' would be constant and small enough
that the possibility of measurement error would be relatively large. Forthese reasons, values of after-mix loss are obtained analytically with the
use of either experimentally or analytically determined before-mix(station 1) loss parameters.
The basic equations for determining the after-mix conditions are
those for conservation of mass, momentum in the tangential direction,
and momentum in the axial direction during mixing. Equating the mass
flow rate before mixing (station 1) and after mixing (station 2) yields
fo (pV)l cos d =cos a2(pV)2(7-74)
From conservation of momentum in the tangential direction we get
f0 (pV2)t at at a_ cos a2 (pV2)2 (7-75)=sin
and from conservation of momentum in the axial direction we get
f 1 (:)a 4 ptd q- (pV 2) costard (7-76)
Although these equations are subscripted for two-dimensional flow, they
can also be applied to three-dimensional flow by integrating radially.
If experimental survey data were available at station 1, the integrals in
the above equations could be directly evaluated even with variations in
static pressure and flow angle. These conservation equations could be
written for any before-mix location at which data were available, and then
used to evaluate the after-mix loss coefficient. In the case where the
before-mix station is not station 1, it would not be possible to determine
114
BOUNDARY-LAYER LOSSES
the mixing loss completely by experimental means. In most cases, how-
ever, it is only the final after-mix loss that is desired, and survey measure-
ments are usually made a little farther downstream of the trailing edge,where angle and pressure variations have somewhat damped out.
If static pressure and flow angle are constant across station 1, it is
possible to express equations (7-74) to (7-76) in terms of the previously
used boundary-layer parameters, as was done in reference 1. The analysisherein differs from that of reference 1 only in that the before-mix station
used in reference 1 corresponds to station la herein. Equation (7-28)
subscripted for station 1, where there is no trailing-edge void, can bewritten as
,fo (pY)Id =(1-_l*)(pV)/,n (7-77)
Subscripting equation (7-29) for station I and combining it with equation(7-77) yields
,fo(pV2)l (1- _l*--0x*) (pVZ)/°.l (7-78)
Substituting equations (7-77) and (7-78) into equations (7-74) to(7-76) yields the following equations for conservation of mass and
momentum in terms of the boundary-layer parameters previously deter-mined:
cos al (1 - _1") (pV) Io,1 = cos a2 (pV) 2 (7-79)
sin al cos al(1--Sl*--01*) (pVS)1,.1=sin as cos as(pVS)s (7-80)
Opt+cos s al(1-_*-0_*) (pVS) i,,1 = cos s as(pV2)s+gps (7-81)
These equations, along with the ideal gas law and the conservation-of-
energy equation (Ta'= Ts'), can be solved simultaneously as shown in
reference 1 to obtain Ss, the after-mix kinetic-energy loss coefficient,for both compressible and incompressible flow.
For incompressible flow, the solution for _ is
sin s al (1 - _I*- 01"\ s__$--_ /'+cos a,(1-_,*)'
_--1 l+2cosSal[_(l__,)2_(l__, 8,)_ ] (7-82)
For compressible flow, no explicit solution was found, and the followingsteps are required to obtain Ss:
115
TURBINE DESIGN AND APPLICATION
(1) The parameters C and D are computed from
3"-I-1 V _(1- A,.n) --_-+cos' aff l-6,*-Ox*) (_-_.),..,
C-V
e°s at(1 --_1") (G)f., 1
V sin alD= V,, 1,.1 \ 1--61" ]
(2) The quantity (V,/V¢r)2 is obtained from
(3) The density ratio (p/p')2 is obtained from
3,-1
(4) The total pressure ratio p2'/po' is obtained from
(7-83)
(7-84)
(7-85)
(7-86)
p0 _
cos al(1 --_1")p Vc, l,,t
pV,,'V\p ,r/2
(7-87)
(5) The pressure ratio (p/p')2 is obtained from
P P(;),(6) Finally, $2 is obtained from
(7-88)
P°'_ (_-_)/_' 1ps-;]
6-
po'_ (_-t)/_kp2 /
(7-89)
Values of 6 include all the blade-row loss; that is, the frictional loss of
the blade row, the trailing-edge loss, and the mixing loss. Values of $1
116
BOUNDARY-LAYER LOSSES
include all the blade-row losses except mixing loss. Therefore,
_._, = _- _ (7-90)
where Sm_ is the fractional loss in available energy due to mixing.
BLADE-ROW LOSS CHARACTERISTICS
In this section, experimentally and analytically determined losses of
the various types considered will be presented and compared, and the
effect of blade-row geometry on losses will be discussed.
Distribution and Comparison of Losses
Figure 7-8, taken from reference 8, compares experimentally and
analytically determined values of kinetic-energy loss coefficients at three
different angle settings for a given stator and at three stations repre-
senting different losses. The loss coefficient _x.... obtained just within theblade trailing edge, represents the surface-friction loss at the mean
(arithmetic mean radius) section; the coefficient $1.,, obtained just
beyond the trailing edge, represents the friction loss plus trailing-edge lossat the mean section; and the coefficient _.3D represents the total loss for
the annulus including blade and end-wall friction, trailing-edge drag, and
mixing. In general, agreement between the experimental and analytical
loss eoefiicients is reasonably good.
.0_
.0]
(
-_ .o2
¢..
_ 0 Experimental results
[2 Analytical results
2,30 Mixin_and
""""_- _'_'1_ end-wall lossesel, m Tralli_-
edgeloss...-------'_ -_ ela, m
_ Mean-section
bladesurface! friction loss
0 I I JlO I_ DO
Percentstatorateasettlno
F1Oua_. 7-8.--Comparison of experimental and analytical loss coefficients for different
stator ares settings. (Data from ref. 8.)
117
TURBINE DESIGN AND APPLICATION
0 _2,3D
e_l,30
.05- _ el,m
r"] ela ' m
I® .'0403__u
_8 Mi xi ng loss
I loss
[ ,. _.02
"" .01
0.5 .6 .7 .8 .9
Mun-sectlonidealafter-mixcriticalvelocity,(VNcr)i,m,2
FIOURE 7-9.--Variation of loss coefficients with velocity. (l)ata from ref. 9.)
Figure 7-8 gives some idea of the distribution of losses in a stator blade
row, but does not separate the mixing and end-wall losses. Figure 7-9,taken from reference 9, shows the variation in loss coefficient with ve-
locity. Loss coefficient is seen to decrease slightly with increasing velocity.This figure also shows separately the mixing and end-wall losses, as wellas the other blade-row losses.
In this particular case, the friction loss was about 2 percent of the stator
ideal energy and about one-half of the total stator loss. The trailing-edge
loss was about one-quarter of the total loss. In general, the trailing-edge
loss will vary with trailing-edge blockage as was shown in figure 7-7.
The end-wall loss, which was about 15 percent of the total loss for thisease, will vary with the design, depending primarily on radius ratio and
spacing. The mixing loss made up the remaining 10 percent of the total
loss. The loss breakdown will, of course, vary with the stator design, butthe comparison does indicate that each of the losses may be of con-
sequence.
Effect of Blade-Row Geometry on Losses
A study of the effect of turbine geometry on turbulent-flow boundary-
layer loss is presented in reference 10. In that study, the assumption was
118
BOUNDARY-LAYER LOSSES
made that the momentum loss per unit blade surface varies as the inverse
of the chord Reynolds number to the m power:
_tot
-- _: ReT" (7-91 )C
where Re¢ is a Reynolds number based on blade chord c. Expanding
equation (7-91) by multiplying and dividing by like terms, substituting
into equation (7-22), and then using equation (7-47) to express the
three-dimensional effect yielded an equation of the form
,,; Fl+cos..l (7-92)
,(-:)where Reh is a Reynolds number based on blade height h. As indicated,
the three-dimensional momentum thickness parameter can be expressed
as a function of the geometric variables--height-to-spacing ratio h/s,blade solidity c/s, and height Reynolds number Reh. The reference value
of OtoJC, as explained in reference 9, is based on the minimum loss for a
given solidity and, therefore, becomes a function of solidity. The ex-
ponent m is set equal to _j in the analysis.
In reference 10, the derivative of the dimensionless momentum thick-
ness 0a* with respect to each of the geometric variables was obtained inorder to find the minimum-loss value of each variable in terms of the other
variables (there is no minimum for height Reynolds number). With the
optimum values known, the relative variations in momentum loss around
the minimum values were then determined. The results of the analysis
from reference 10 are shown in figures 7-10, 7-11, and 7-12. Also shown in
each figure is the nature of the geometry variation associated with the
change in each variable.
Figure 7-10 shows that a wide variation (50 percent or more) in h/s
value around the optimum causes little increase in momentum loss. This
results from the two counteracting effects of changes in chord Reynolds
number and end-wall area. Figure 7-11 shows that the solidity of a blade
may be varied considerably from optimum with some, but not excessive,
loss. Comparison of the results in figures 7-10 and 7-11 shows that the
loss is more sensitive to solidity than to the height-to-spacing ratio. The
curve shape of figure 7-11 reflects also the counteracting influences of
chord Reynolds number and end-wall area.
119
TURBINE DESIGN AND APPLICATION
_ =1.4--
_ 1.o.2 .6
"/4/////: ////(t?
/A "////,,///[//
/N//i/
J/,,;,"
cls and Re h are constant
I I _-----r---_lLO 1.4 1.8 Z2 Z6
Height-to-spacingratiorelativetooptimumratio,th/s)/(h/s)opt
FIGURE 7-10.--Variation of momentum-thickness ratio with variation in height-to-
spacing ratio. (Data from ref. 10.)
//"l/IlL I///{/!!I!
"/,/////2 "/_[/
I
.:'' ¢
2'c 1.4-- _/
_*_ his and Reh areco
_$ Lo I b'.-_ I J-._ I I.4 .6 .8 LO 1.2 L4 1.6
Solidity ratio, (cls)llcls)op t
FIGURZ 7-11.--Variation of momentum-thickness ratio with solidity ratio. (Data
from ref. 10.)
120
BOUNDARY-LAYER LOSSEF
2.0_
'1.8
_i 1"6
1.4
1.2E
E 1.0--
.8 I I I I0 .2 .4 .6 .8
'.,'(.Z__//_//-Z/-._I
I
1I
ill _/, ,///_/
2/
/'
e'." /
j,his and cls are constant
]111 I--7LO L2 L4 L6 L8 ZO
HeightReynoldsnumber ratio, RehlReh,ref
FIOURE 7-12.--Variation of momentum-thickness ratio with height Reynolds number
ratio. (Data from ref. 10.)
Figure 7-12 shows the variation of momentum-thickness ratio with
height Reynolds number ratio. While the figure indicates a change in
Reynolds number due to change in geometry, the change in Reynolds
number could also result from change in inlet flow conditions. The curve
shape, then, results from the loss being inversely proportional to the
Reynolds number to the m=_ power. These results show that an in-
crease in height Reynolds number results in improved performance. The
height Reynolds number is sometimes used in correlating the performance
of different turbomachines.
REFERENCES
1. STEWART, WARNER, L.: Analysis of Two-Dimensional Compressible-Flow LossCharacteristics Downstream of Turbomachine Blade Rows in Terms of Basic
Boundary-Layer Characteristics. NACA TN 3515, 1955.
2. McNALLY, WILLIAM I).: FORTRAN Program for Calculating CompressibleLaminar and Turbulent Boundary Layers in Arbitrary Pressure Gradients.
NASA TN D-5681, 1970.3. PATANKAR, S. V.; ANn SPALDING, D. B.: Heat and Mass Transfer in Boundary
Layers. CRC Press, 1967.
4. WHITNEY, WARREN J.; STEWART, WARNER L.; AND MISER, JAMES W.: Experi-
mental Investigation of Turbine Stator-Blade-Outlet Boundary-Layer Char-acteristics and a Comparison with Theoretical Results. NACA RM E55K24,1956.
121
TURBINE DESIGN AND APPLICATION
5. STEWART, WARNER L.; WHITNEY, WARREN J.; AND WONG, ROBERT Y.: Use of
Mean-Section Boundary-Layer Parameters in Predicting Three-Dimensional
Turbine Stator Losses. NACA RM E55L12a, 1956.
6. HOEHNER, SIGHARD F.: Fluid-Dynamic Drag. Midland Park, N.J., 1965.
7. PRUST, HERMAN W., JR.; AND HELON, RONALD M. : Effect of Trailing-Edge Geom-
etry and Thickness on the Performance of Certain Turbine Stator Blading.NASA TN 1)-6637, 1972.
8. PRUST, HERMAN W.; MOFF1TT, THOMAS P.; AND BIDER, BERNARD: Effect of
Variable Stator Area on Performance of a Single-Stage Turbine Suitable for
Air Cooling. V--Stator Detailed Losses with 70-Percent Design Area. NASATM X-1696, 1968.
9. MOFFITT, THOMAS P.; PRUST, HERMAN W., JR.; AND BIDER, BERNARD; Effect of
Variable Stator Area on Performance of a Single-Stage Turbine Suitable for
Air Cooling. II--Stator Detailed Losses with 130-Percent Design Area. NASATM X-1635, 1968.
10. MISER, JAMES W.; STEWART, WARNER L.; AND WHITNEY, WARREN J.: Analysis
of Turbomachine Viscous Losses Affected by Changes in Blade Geometry.NACA RM E56F21, 1956.
122
BOUNDARY-LAYER LOSSES
SYMBOLS
A, surface area of one blade, m2; ft 2
AIo parameter defined by equation (7-15)
A. surface area of end walls for one passage, m2; ft 2
C parameter defined by equation (7-83)C_ drag coefficient
c blade chord, m; ft
cp specific heat at constant pressure, J/(kg) (K) ; Btu/(lb) (°R)
[drag, N/m; lb/ftD [parameter defined by equation (7-84)
E energy factorkinetic-energy loss coefficient
0 conversion constant, 1 ; 32.17 (Ibm) (ft) / (lbf) (sec 2)H form factor
[blade height, m; fth _height of body placed in boundary layer, m; ft
J conversion constant, 1; 778 (ft) (lb)/Btu
l blade surface distance from forward to rear stagnation point,
m; ft
m exponent in equation (7-91)
n turbulent boundary-layer velocity profile exponent
p absolute pressure, N/m2; lb/ft _
q dynamic pressure, N/mS; lb/ft _
Re_ chord Reynolds numberRe_ height Reynolds number
r radius, m; ft
s blade spacing, m; f_
T absolute temperature, K; °R
t trailing-edge thickness, m; ft
u distance in tangential direction, m; ftV fluid velocity, m/sec; ft/sec
x distance along blade surface from forward stagnation point,
m; ft
Y distance from surface normal to boundary layer, m; ft
y distance from surface normal to boundary layer expressed as
fraction of boundary-layer thickness
a fluid flow angle from axial direction, deg
a, blade staggerangle from axial direction, deg
-y ratio of specific heat at constant pressure to specific heat atconstant volume
5 boundary-layer displacement thickness, m; ft
_/., boundary-layer thickness, m; ft
123
TURBINE DESIGN AND APPLICATION
boundary-layer momentum thickness, m; ft
viscosity, (N) (sec)/m2; lb/(ft) (sec)
density, kg/m_; lb/ft 3
boundary-layer energy thickness, m ; ft
Subscripts:
cr critical flow conditions
eft effective
f friction
fs free streamh hub
i ideal
inc incompressiblem mean section
rain minimum
mix mixing
opt optimum
p pressure surface
ref references suction surface
t tip
te trailing edgetot total
x axial component0 blade-row inlet
1 just beyond trailing edge of blade row
la just within trailing edge of blade row2 downstream uniform state
3D three dimensional
Superscripts:
absolute total state
* dimensionless value
124
CHAPTER8
MiscellaneousLosses
By RichardJ. Roelke
In the last chapter, the boundary-layer losses associated with the flow
process in the blade channel were discussed. To determine the overall
design-point efficiency of a turbine, other losses must also be considered;these include tip-clearance loss and disk-friction loss. In some instances,
these losses represent a very small part of the turbine output and may be
neglected; however, in other instances, these losses can be of such mag-nitude as to influence the selection of the turbine design point. The sum
of these losses normally comprises all the losses that are considered in the
design of a full-admission axial-flow turbine. If, however, a partial-
admission turbine is being considered, there are additional losses that
must be included. The partial-admission losses usually considered are the
pumping loss in the inactive blade channels and the filling-and-emptying
loss in the blade passages as they pass through the admission arc. Finally,
a loss that occurs at off-design operation of any turbine is the incidence
loss, which will also be covered herein.
TIP-CLEARANCE LOSS
Because a turbine must operate with some clearance between the tips
of the rotor blades and the casing, some fraction of the fluid leaks across
the tips, thus causing a reduction in turbine work output. This leakage
loss is affected, first of all, by the nature of the tip geometry; that is, by
the amount of radial clearance, by recesses in the casing, and by tip
shrouds. For a given tip geometry, the amount of blade reaction affects
the leakage loss, since a large pressure difference across the tip (high
reaction) causes more higher-kinetic-energy flow to leak through the tip
125
TURBINE DESIGN AND APPLICATION
gap from the pressure side to the suction side of the blade. With an un-
shrouded blade, this leakage flow not only causes a loss due to its own
reduced work, but also causes an unloading of the blade, primarily in the
tip region. Analytical evaluation of the drop in turbine efficiency caused
by tip-clearance leakage is inherently difficult because of the complex flowproblem. Several empirical expressions for clearance loss have been
developed, and some of these are summarized in reference 1; however,
they are rather complicated, and the author states that none is entirelysatisfactory.
A number of tests have been made at the NASA Lewis Research Center
to determine the effect of tip clearance and tip geometry on axial-flowimpulse and reaction turbines. An examination of some of the results of
these tests helps to obtain a better understanding of the tip-clearance
loss. Figure 8-1 shows the angle traces at the blade exit of a 5-inch single-
stage turbine (ref. 2). Two things to be noted from the angle traces are
that the flow in the clearance space and near the tip was not fully turned,
i+
Ro_lon_
Tipclearance,percentof
pessageheight Axl ,_ direction"0 1.2
A 3.1o ,0o if-
i Innerwall OuterwallIl I I J i.6 .7 .8 .9 1.0
Ratioof hubrNlus totipradius
FIovRz 8--1.--Variation of exit flow angle with raditmratio for four rotor tip clearances.(Data from ref. 2.)
126
MISCELLANEOUS LOSSES
J Turbine
J 0 1-Stage, reaction (ref. 2)
I [D 2-Stage, reaction (ref. 3)
Z_ l-Stage, impulse (ref. 4)
J - - - Estimate (ref. 5) for ref. 2 turbine
l ----Estlmate(ref. 5)for ref. 4turbine
•
,o
!- al.g'._ .N
.76 I I I I I I0 .0_ .04 .06 .08 .10 .12
Tip clearance, fractinn n_ ,, _ =no hoinht
FIGURE 8-2.--Effect of tip clearance on efficiency.
even at the smallest clearance tested, and that underturning of the flowincreased with increasing tip clearance, and this effect occurred all the
way down to the hub. This underturning of the flow unloads the blade
aerodynamically and results in lower turbine output and efficiency. The
decrease in efficiency for this turbine, as well as for two others, is shownin figure 8-2.
The solid lines in figure 8-2 represent test results from single-stage
(ref. 2) and two-stage (ref. 3) reaction turbines and from a single-stage
impulse turbine (ref. 4). All turbines were unshrouded. The importancethat the level of reaction plays in the clearance loss is clearly evident from
the figure. For the same ratio of tip clearance to blade height, the losses in
efficiency for the reaction turbines were about double that for the impulseturbine.
The dashed lines in figure 8-2 are estimates of the efficiency losses forthe two single-stage turbines (refs. 2 and 4) as obtained from the curves
published in reference 5 (as fig. 1.6) and reproduced here as figure 8-3.
Extrapolation of the experimental data of figure 8-2 shows that figure 8-3gives satisfactory estimates of tip-leakage loss for small tip clearances.
127
TURBINE DESIGN AND APPLICATION
1.00
t_
co
.u
F--
I I.01 .02
Tip clearance, fraction of passageheight
Fmua_. 8-3.--Tip-clearance correlation for unshrouded blades• (Data from ref. 5.)
Reviewing the results shown in figures 8-2 and 8-3, it is apparent that the
loss in efficiency increases with increasing reaction, and for moderate to
large ratios of tip clearance to blade height, the loss is appreciable.
In addition to reducing the tip clearance, methods for reducing the tip-
leakage losses include recessing the casing above the blade tip while in-
creasing the blade height, and adding a tip shroud. These loss-reduction
schemes can be used either individually or in combination. The single-
stage impulse turbine of reference 4 was tested at several ratios of tip
clearance to blade height, both without and with the recessed casing and
the tip shroud. Figure 8-4 shows the three general configurations tested
in reference 4, and the turbine-performance results are shown in figure 8-5.
A clearer understanding of the performance characteristics is possible if
the loss mechanisms are considered. The factors affecting turbine work
for the reduced blade-height configuration as compared to a zero-clearance
configuration consist of (1) reduced blade loading area, (2) clearance-gap
leakage flow, (3) mixing of the leakage flow with channel throughflow,
128
Flow
(a)
MISCELLANEOUS LOSSES
Recessedcasing_
I
Flow
blade
(b)
Tip shroud-_
',-Rotor
blade
(c)
(a) Reduced blade height (relative to (b) Recessed casing.
zero-clearance blade height).
(c) Shrouded rotor.
_FIOURE 8-4.--Tip-clearance configurations investigated for impulse turbine (ref. 4).
and (4) blade unloading (as a result of flow going from the pressure side
to the suction side). With the recessed-casing configuration, the blade
extended to the passage outer radius and was of constant height as the
clearance gap was changed by varying the amount of casing recess. There-
fore, the reduced blade loading area was eliminated, and the leakage flowwas reduced because of the indirect leakage path. With the shroud added
to the blade, the blade unloading was eliminated, and the leakage flow was
further reduced. Note from figure 8-5, however, that at tip-clearance
129
TURBINE DESIGN AND APPLICATION
m
o o
&
8
J_
38--
o Shroudedrotor
1 I I I 1• _ .04 .06 .08 .10 -.12
Rotortipclearance,fractionofpassageheight
FIGURE 8-5.--Effect of tip-clearance configurations on turbine efficiency. (Data fromref. 4.)
fractions below some value, about 0.035 in this instance, the shroud no
longer provides an increase in efficiency. This can be attributed to an
increasing friction loss between shroud and casing as the clearance gapis decreased.
The comparative results shown in figure 8-5 for different blade-tipgeometries are dependent upon that particular design and may not apply
to other turbines. This is particularly true of the shrouded blade, since
the leakage flow depends not only on the clearance span and pressuredifference but also on the number of seals used. With respect to the
recessed-casing configuration, it should be noted that the blade should not
extend into the recess. If it does, the overlapping section will just bechurning stagnant fluid and creating additional losses.
In summary, tip-clearance loss presents a complicated flow probleminfluenced by many factors and is not easily predicted with consistent
accuracy. The clearance gap required for a turbine depends primarily on
diameter (larger clearance for larger diameter) and, as seen previously,the loss increases as the ratio of clearance gap to blade height increases.
130
MISCELLANEOUS LOSSES
For any given diameter, therefore, the tip-clearance loss increases with
increasing hub- to tip-radius ratio. It becomes increasingly difficult to
maintain a desired small ratio of clearance gap to blade height as the
turbine, and hence the blade height, becomes smaller. For a given radius
ratio, therefore, the loss is more severe for small turbines and less severe
for larger turbines. If tip leakage is considered to be a problem in a
particular case, it might be worthwhile to carry out tests to evaluate the
leakage effects.
DISK-FRICTION LOSS
The disk-friction loss (or windage loss) is due to the skin friction and
circulation of fluid between the rotating disk and the stationary casing. In
addition, some turbines for hot applications, for example aircraft engines,
have a small steady stream of lower-temperature gas that bathes and cools
the rotor disk. This cooling gas flows along the rotor-disk surface from
near the engine centerline outward to the base of the blades. The qualita-tive nature of the flow patterns around rotor disks without and with
throughflow of cooling gas are shown in figure 8-6. Equations for es-
timating the associated losses are presented herein.
No Throughflow
For the case with no throughflow, as in figure 8-6(a), the thin layer offluid close to the rotating surface is thrown outward by centrifugal action
(a) lz!////_//!I
T
(b)
(a) Without throughflow. (b) With throughflow.FIGURE 8-6.--Flow patterns for rotating disks.
131
TURBINE DESIGN AND APPLICATION
and returns via the stationary wall to the inner radius, thereby building
up a continuous circulatory effect. Consider an element of area on oneside of the disk
dA = 27rr dr (8-1)
where A is the area, in m _or ft 2, of one side of the disk, and r is the radius,
in m or ft, of the area element dA. The fluid shear stress r, in N/m _ or
lb/ft 2, acting over this area at the radius r produces a resisting torque tothe disk rotation of
dMo-- = r2rr _ dr (8-2)
2
where Mo is the resisting torque, in N-m or lb-ft, for both sides of a disk
in the case of no throughflow. The shear stress can be expressed as
where
CI
g
P
Vu
C$
r=_g pVJ (8-3)
fluid shear-stress coefficient
conversion constant, 1; 32.17 (lbm) (ft)/(lbf) (sec 2)
density, kg/m3; lb/ft '_tangential component of fluid absolute velocity, m/sec; ft/sec
At the disk surface, the fluid tangential velocity is
V_ = r_ (8-4)
where w is the angular velocity, in rad/sec. By substituting equation (8--4)
into (8-3), the total torque for both sides of the disk can be written as
"2
Mo= fo g Cypo_2r4dr (8-5)
where a is the disk rim radius, in m or ft. Performing the integration yields
w2a 5Mo = CM.oP -- (8--6)
2g
where C.,o is a torque coefficient for the case of no throughflow. The disk-
friction loss expressed as power is then the torque times the angular
velocity:
AioW CM,o Pw'_a5 (8-7)Pdl =-7 = 2gJ
where Pdy is the disk-friction power loss, in W or Btu/sec, and J is a
132
MISCELLANEOUS LOSSES
conversion constant (equal to 1, or 778 (ft)(lb)/Btu). The form of
equation (8-7) that is found in most handbooks is
Paj = KdlpNaD 5 (8-8)
where
K_I disk-friction power-loss coefficient
N rotative speed
Dr disk rim diameter
A number of investigators have published values of the constant K_f in
equation (8-8) to be used for different circumstances, while others have
made small changes to the exponents to better fit the available data. The
wide assortment of semiempirical equations used to predict this loss is, no
doubt, due to variations of the test-apparatus geometry, the somewhat
oversimplified model from which equation (8-7) is derived, and the
existence of different types of flow that can occur in the space between the
rotor and the casing. One thing that can be noted from equation (8-7)
or (8-8) is that for a given blade speed, lower loss is obtained by having
a smaller diameter and a higher rotative speed.
An extensive investigation has been conducted (refs. 6 and 7) to deter-
mine the effect of chamber proportions on disk friction and to present a
clearer picture of the several modes of flow that may exist. In general,four modes of flow, or flow regimes, can exist in the axial space between the
casing and the rotating disk, depending on the chamber dimensions and
the flow Reynolds number. The torque coefficient CM.o was evaluated
both theoretically and experimentally in each regime. A description of
each regime and the associated equations for the torque coefficient are asfollows:
Regime I: Laminar Flow, Small Clearance. Boundary layers on therotor disk and casing are merged, so that a continuous variation in
velocity exists across the axial gap s. Figure 8-7 (a) indicates the natureof the variations in the radial and tangential components of fluid velocity
at a given radius in the gap. The best equation for torque coefficient, both
theoretically and empirically, is
2_ (s-9)CM.o-- (s/a) R
where s is the axial distance, in m or ft, between disk and casing, and R is
the Reynolds number defined as
R =_a2p (8-10)
where _ is the dynamic viscosity, in (N) (sec)/m _or lb/(ft) (sec).
133
TURBINE DESIGN AND APPLICATION
Radial Ta_entlalcomponent componentof velocity of velocity
0 /t°
///////////7/I/I///ii/,_
r_
(a)
Radialcomponentofvelocity
Tangentialcomponentofvelocity
0
(b)
(a) How regimes I and Ill.
(b) How regimes II and IV.
FmuaF, 8-7.--Velocity patterns around rotating disks without throughflow.
Regime I I: Laminar Flow, Large Clearance. The combined thick-
ness of the boundary layers on the rotor and on the casing is less than the
axial gap, and between these boundary layers there exists a core of rot aN
ing fluid in which no change in velocity occurs. Figure 8-7(b) shows
the variations in the radial and tangential velocity components for this
case. The best theoretical and empirical equations for torque coefficientare
VII_q
C.v,o R1/2 (8-11)
where CII is a function of (s/a), as shown in figure 8-8(a), and
134
MISCELLANEOUS LOSSES
E
¢.J
3.3 m
3'1 i
2.9
2.7
(a)
I I I I I2.5
• 10
._.0_
•07
(b)
I I [ I I•05 .10 .15 .20 .25
Ratioof axial gapto rim radius, s/a
I.30
(a) Flow regime II.(b) Flow regime IV.
FIGURE 8-8.--Evaluation of torque coefficients. (Data from ref. 6.)
CM,o-- 3.70 (s/a)_/1o (8-12)RII2
respectively.
Regime III: Turbulent Flow, Small Clearance• The turbulent
counterpart of Regime I. The best theoretical and empirical equations
for torque coefficient are
0.0622
CM,o-- (s/a) '/4R'/_ (8-13)
and
0.O8O
CM,o= (s/a) l/eR1/4 (8-14)
respectively.
135
TURBINE DESIGN AND APPLICATION
Regime IV: Turbulent Flow, Large Clearance. The turbulent
counterpart of Regime II. The best theoretical and empirical equa-tions for torque coefficient are
C Cxv
u '° = -R-f-A (8-15)
where Czv is a function of (s/a), as shown in figure 8-8 (b), and
CM,o = O. 102 (s/a) ,/10R1/_ (8--16)
respectively.
The particular flow regime that exists at any Reynolds number can be
determined by plotting torque coefficient against Reynolds number from
equations (8--9), (8-11), (8--13), and (8-15), as shown in figure 8-9 for
several values of s/a. The discontinuities (changes ill slope) in the lines
of figure 8-9 indicate transition from one regime to another. In this figure,
the flow regimes are determined by matching the slopes of the lines with
Slopeof curve Flow I Descriptionreg,mej
I Laminarflow; mergedboundarylayers
II Laminar flow; separateboundary layers
III Turbulent flow; mergedboundarylayers
IV Turbulent flow; separate,.. boundarylayers
Ratioof axial'gapto disk rim radius,
s/a
----- 0.01.02.05.20
F1GURE 8-9.--Theoretical variation of torque coefficient with Reynolds number for nothroughflow. (Data from ref. 6.)
136
MISCELLANEOUS LOSSES
those shown in the insert in the figure. Torque-coefficient values (ref. 6)determined experimentally with a 50.8-centimeter (20-in.) disk rotated in
water and oil for several values of s/a verify the theory.
Throughflow
For the case of the rotating disk with throughflow, as in figure 8-6(b),
the friction torque increases with the throughflow. This problem has been
analyzed for low values of throughflow with regime-IV flow. In this case,it is assumed that the fluid enters the chamber near the centerline with no
angular velocity and leaves at the rim with some angular velocity K,_a.
The symbol Ko represents the ratio of the angular velocity of the rotating
core of gas to the angular velocity of the disk. The increase in torque,
AM, over that without throughflow is the rate of change of angularmomentum of the fluid flowing through the system:
hM=2p Q- (gg_a)a=2p Q gj_a _ (8-17)9 9
where Q is the volumetric throughflow rate, in m_/sec or ft3/sec, in the
clearance space on one side of the disk. The total torque for the through-flow case is then
M=Mo+AM CM'°p_2aS+2 P-QK_a 2 (8-18)2g g
The value of Ko is approximately 0.45 for s/a ratios from 0.025 to 0.12.
An assessment of the power loss can be obtained by calculating thefriction torque of the throughflow case compared to that of the no-throughflow case:
.4Ko QM__._= 1+ 2pog_'°a_ - 1_M_M.o _a 3 (8-19)Mo 1
"_C M,op_a 5
Substituting equation (8-16) for CM.o yields
M K°R1/5 -_Q.= 1+39.2 KoM---_= 1+0.0255 (s/a)mo _a" (s/a) mo-- v (s-2o)
where T is a dimensionless throughflow number defined as
OT = _ R 1/5 (8--21)
oja 3
According to the data of reference 7, equation (8--20) predicts values
that are somewhat high; moreover, the effect of s/a is not accurately
137
TURBINE DESIGN AND APPLICATION
'1.4
1.3
£
_1.2
o
1.1
1.00
Ratioofaxial gaptodisk rim radius,
sla
_0. 03.06
.01 ._ .03 .04 .05Throughflownumber, T
Fmua_ 8-10.--Empirical variation of torque with throughflow number. (Data from
ref. 7.)
given by (s/a) _/L°. Empirically, the test data are represented to within
±5 percent by the relation
M T
i---: = 1 +13.9K0 (s/a)z/-----_ (8-22)
Equation (8-22) is plotted in figure 8-10 for several s/a values.
PARTIAL-ADMISSION LOSSES
Full-admission axial-flow turbines are used for most applications;
however, unusual conditions sometimes arise for which a partial-admission
turbine may be a better choice. If, for example, the design mass-flow rate is
so small that a normal full-admission design would give very-small blade
heights, then it may be advantageous to use partial admission. The losses
due to partial admission with long blades may be less than the leakage
and low Reynolds-number losses of the full-admission turbine having short
blades. In addition, for a given rotative speed, partial admission allows the
freedom of larger diameter and higher blade-jet speed ratios. Also, the
use of partial admission may be a convenient way to reduce power output
138
MISCELLANEOUS LOSSES
of an existing full-admission turbine (physically block some of the stator
passages). In general, partial-admission turbines have high specific-work
output and low volumetric-flow rates.
As mentioned previously in this chapter, the partial-admission losses
are the pumping loss in the inactive blade channels and the filling-and-
emptying loss encountered as the blades pass through the active sector.
This latter loss has been referred to as expansion, scavenging, or sector
loss. The mechanisms of partial-admission losses are not clearly or fully
understood, but they do result in a decrease in output power and efficiency
when compared to the same turbine operating with full admission.
The pumping loss is that loss caused by the inactive blades rotating in
a fluid-filled casing, and expressions for it are somewhat similar in form to,and often combined with, the expression for the disk-friction loss. These
expressions all seem to trace back to reference 8, where the results of
several experimental investigations are summarized. The equations for
estimating pumping-power loss that resulted from these investigations
showed that the effects of blade height and diameter on the pumping-
power loss are quite uncertain, as evidenced by variations in the exponents
on these terms. Further, the nature and location of obstructions (adjacent
blade rows, casing wall, etc.) or lack of such in the vicinity of the three
open sides of the blade channel were accounted for only by differences in
the empirical loss coefficient. Therefore, it appears that a generallyapplicable expression for pumping-power loss is yet to be found.
The one equation perhaps most often used is
where
Pp
K_U,
l
D.
E
PI, = K_,pU,,,Sl I'BD- ( 1 - _) (8-23)
pumping-power loss, W; (ft) (lb)/sec
pumping-power loss coefficient, 1/m'/_; (lbf) (sec _)/ (Ibm) (ft 3/_)
blade mean-section speed, m/sec; ft/secblade height, m; ft
blade mean-section diameter, m; ftactive fraction of stator-exit area
The value of the coefficient Kp as reported in reference 8 and converted
to the units used herein is 3.63 l/m 1I_, or 0.0105 (lbf) (sec2)/(lbm) (fta/2),
for an unenclosed rotor. For the same rotors enclosed, the coefficient
values were one-quarter to one-half of the above values. More recently,the combined disk-friction and pumping losses for a single-stage rotor
enclosed by the turbine housing were reported in reference 9. If a disk-
friction loss estimated by equation (8-7) is subtracted from the combined
losses of reference 9 and if the remaining loss is converted to the form of
equation (8-23), the coefficient Kp is found to be 5.92 1/m 1/_, or 0.0171
(lbf) (sec _) /(lbm) (ft_/_). This is significantly higher than the coefficients
139
TURBINE DESIGN AND APPLICATION
reported in reference 8, and the difference is attributed to primarily the
lack of an adequate loss model.
The other partial-admission loss shall herein be called the sector loss.
Imagine a blade channel as it just starts to enter the active sector. It is
filled with relatively stagnant fluid that must be pushed out by the high-momentum fluid leaving the nozzle. This scavenging will continue until
the blade channel is completely within the active sector. As the blade
channel passes out of the active sector, a second sector loss occurs. As theinlet to the blade channel is cut off from the nozzle active arc, less and
less high-momentum fluid enters the channel. Since this fluid has the
entire blade channel area to flow into, it is rapidly diffused as it flows
through the rotor. These losses cause an overall decrease in the momentumof the fluid passing through the rotor, thus decreasing the available energy
of the fluid. It was reported in reference 10 that this decrease in momen-
tum may be found by multiplying the rotor-exit momentum by a losscoefficient
K,=(l-_f) (8-24)
where p is the rotor-blade pitch, in m or ft, and f is the nozzle active arc
length, in m or ft. Effectively, K, is a rotor velocity coefficient that ac-counts for the sector loss.
The effect of the sector loss on turbine efficiency is determined as
follows. With the use of equations (2-6) and (2-14), from volume 1, and
the associated velocity diagram geometry, we can express the specific
work of an axial-flow turbine as
z_h'= (W_ a- W_,,) -- -_ (W, sin B_- W, sin B,) (8-25)
where
W_W
turbine specific work, J/kg; Btu/lb
tangential component of relative velocity, m/sec; ft/sec
relative velocity, m/see; ft/sec
fluid relative angle measured from axial direction, deg
The subscripts 1 and 2 refer to the rotor inlet and exit, respectively. For
an impulse turbine (which most partial-admission turbines are), where
,_, =U., W1 sin _I(1-t-K_) (8-26)gJ
where K. is the rotor relative-velocity ratio W_/WI for the full-admission
turbine. For the partial-admission turbine, applying the sector loss
140
MISCELLANEOUS LOSSES
coefficient yields
W2 ffiKwK,W1
So, for the partial-admission turbine,
Ah'_= U__ W_ sin _1(1 +K_K,)OJ
Since efficicncy is
(8-27)
(s-2s)
(8-30)
into equation (8-30) then
Ah'
= Ahid (8-29)
where Ahld is thc turbine ideal specific work, in J/kg or Btu/lb, the ef-
ficiency of the partial-admission turbine with respect to that of the full-admission turbine is
_, Ah'_a
Ah'
and (8-28)Substituting equations (8-26)
yields
1 +K_K,(8-31)
_=_ I+K_
The efficiency penalty expressed by equation (8-31) accounts for the
sector loss only; the pumping loss discussed earlier will reduce the overall
efficiency further. Equation (8-24) indicates that a partial-admissionturbine rotor should have closely spaced blades to reduce the sector loss;
however, as more blades are added to the rotor, the blade profile loss will
increase. Also, the effect of the number of rotor blades on the pumping
loss is not known. Therefore, the complete optimization of a partial-
admission design cannot be done analytically at present.
In the study of reference 9, the efficiency of a small axial-flow turbine
was determined over a range of admissions from 12 to 100 percent. The
total loss due to partial-admission operation was taken as the difference
between the full- and the partial-admission efficiencies. The blade pumping
and disk-friction losses were measured separately and were subtracted
from the total partial-admission loss to give what was called other partial-
admission losses. These other losses include the sector loss and any loss
due to leakage from the active sector to the inactive sector. The partial-
admission losses of reference 9 are plotted against admission-arc fraction
in figure 8-11. The combined pumping and disk-friction loss increased
with decreasing arc fraction, while the other losses remained nearly con-
stant over the range of arcs tested.
Predicted efficiencies (from ref. 10) are plotted against blade-jet speed
141
TURBINE DESIGN AND APPLICATION
w
,/r
r_
"611_ -- - -- Estimated pumping and disk-friction losses
"5 I1_ 013 OtherPUmplngand disk-friction losses
• 41"_-_ _rtlal-edmisslon losses
.34
I I.1 .2 .3 .4 .5 .6 .7
Active fraction of stator exit area,
FIGURE 8-11.--Variation of partial-admission losses with active fraction of stator
area. (Data from ref. 9.)
.?--
.6
.5
tJ
,_u .3
bq
.2
.1
I0 1.0
ulmss°n- \
.1 .2 .3 .4 .5 .6 .7 .8 .9
Blade-jet speed ratio, Umlt/- _ Ahld
FIGURE 8-12.--Design-point performance of partial- and full-admission turbines.
(Data from ref. 10.)
ratio (see discussion in vol. 1, ch. 2) in figure 8-12 for a particular turbine
operating with full admission and with three different amounts of partial
admission. The expected reduction in peak efficiency with reduced arc of
admission is seen. The important thing to note from this figure is thereduction in optimum blade-jet speed ratio as admission arc is reduced.
Aerodynamic efficiency is a maximum at a blade-jet speed ratio of 0.5,
142
MISCELLANEOUS LOSSES
irrespective of admission arc, and decreases with decreasing blade speed.
Blade-pumping and disk-friction losses, which decrease with decreasing
blade speed, become a larger part of the gross aerodynamic power asadmission arc decreases. Therefore, as admission arc is reduced, the
maximum net output power (aerodynamic power minus blade-pumpingand disk-friction power) is obtained at lower blade speeds. Thus, for the
design of a partial-admission turbine, the partial-admission losses must be
factored int.() the design before an optimum or near-optimum blade-jetspeed ratio can be selected.
INCIDENCE LOSS
The incidence loss is that loss which occurs when the gas enters a blade
row (either stator or rotor) at some angle other than the optimum flow
angle. Flow incidence would normally only occur at off-design conditions,
since, theoretically at least, all gas and blade angles are matched at the
design condition. The nomenclature used when speaking of incidence is
shown in figure 8-13. The dashed line running through the blade profile
is the camber line and defines the blade inlet angle. The incidence angleis defined as
where
i
a
Ctb
i = a-- _ (8-32)
incidence angle, deg
fluid flow angle from axial direction, deg
blade inlet angle from axial direction, dog
The fluid flow angle must be the absolute angle for stators and the relative
angle for rotors. The incidence angle may be positive or negative, as
indicated in figure 8-13. The sign of the incidence angle is importantbecause cascade tests have shown that the variation of loss with incidence
angle is different for positive and negative angles.
Axial
directionIP
%_../ Vn J I-a-a b
FmURE 8-13.--Blade incidence nomenclature.
143
TURBINE DESIGN AND APPLICATION
Low-reaction_/j
blades-../_
. .._.---------- 0 ------_ +
Inciderce angle, i
F]QUaE 8-14.--Characteristics .f blade incidence loss.
FmuaE 8-15.--L.cal flow separation on })lade surface.
The general nature of the variation of incidence loss with incidence angle
is shown by figure 8-14, which represents a summary of cascade test
results. The loss curve is not symmetrical about the zero incidence angle,
but shows a loss that is larger for positive incidence than for negative
incidence. This may be due to some local separation on the suction surface
at large positive incidence, as indicated in figure 8-15, and the lack, orsmaller area, of separation at the same value of negative incidence. Also,blades in which the mean acceleration of the gas flow is large (high-
reaction blades) have a wide range of incidence over which loss is low,
whereas low-reaction blades have higher losses for the same incidence
range.Another thing to be noted from figure 8-14 is that the minimum loss
does not occur at zero incidence, but at some small amount of negative
incidence. This may be explained by the sketch of figure 8-16. The stag-
nation streamlines for two inlet flow angles are shown; one at zero in-
cidence and the other at some small negative incidence with respect to
the blade inlet angle. Both tests and theory show that the stagnation
streamline curves upward as the flow impacts on the blade leading edge,and the true zero incidence occurs when there is some negative incidence
144
MISCELLANEOUS LOSSES
a<Q
%
FIGURI'_ 8-16.--Curvature of stagnation streamline at blade inlet.
relative to the free-stream flow. The incidence angle at minimum loss is
usually -4 ° to -8 °. Because of this, some turbine designers design theirblades with a small amount of negative incidence, while others do notbecause of the small difference involved.
The magnitude of the incidence loss takes on importance when the
off-design performance of a turbine must be predicted. A method for
determining incidence loss based on test data is described in reference 11.
An analytical method is described here with the aid of figure 8-13. The
inlet velocity V_ can be resolved into a component V, normal to, and a
component Vp parallel to the blade inlet direction (camber line at inlet).If it is assumed that the parallel component passes through the blade row
without any entry loss and that the normal component is entirely lost,
the recovered kinetic energy is
Vl' (V y V,,2oS - 2gS \-_x] = _OJ cos' i (8-33)
and the kinetic-energy loss due to incidence is
L_- V12- 2_ (1 - cos 2 i) (8-34)
In order to account for the differences in loss variation with positive
and negative incidence, the effect of blade-row reaction, and the minimum
loss not occurring at zero incidence, equation (8-34) has been generalizedto
L,=_O J I-l-cos" (i-iop,) ] (8--35)
where io_,, is the optimum (minimum-loss) incidence angle. This type of
equation has proved satisfactory when used in off-design performance
145
TURBINE DESIGN AND APPLICATION
prediction methods such as that of reference 12. Where specific incidence-
loss data are lacking, values of n = 2 for negative incidence and n = 3 forpositive incidence have been used satisfactorily.
REFERENCES
1. HORLOCK, JOHN H.: Axial Flow Turbines. Butterworth hlc., 1966.
2. HOLESKI, ])ONALD E.; AND FUTRAL, SAMUEL M., JR.: Effect of Rotor Tip Clear-
ance on the Performance of a 5-Inch Single-Stage Axial-Flow Turbine. NASA
TM X-1757, 1969.
3. KOFSKEY, MILTON G.; AND NUSBAUM, WILLIAM J. : Performance Evaluation of
a Two-Stage Axial-Flow Turbine for Two Values of Tip Clearance. NASA TN
I)-4388, 1968.
4. KOFSKEY, MILTON G.: Experimental Investigation of Three Tip-Clearance Con-
figurations Over a Range of Tip Clearance Using a Single-Stage Turbine of
High Hub- to Tip-Radius Ratio. NASA TM X-472, 1961.
5. HONG, YONG S.; AND GROH, F. G.: Axial Turbine Loss Analysis and Efficiency
Prediction Method. Rep. D4-3220, Boeing Co., Mar. 11, 1966.
6. DAILY, J. W.; AND NECE, R. E.: Chamber l)imension Effects on Induced Flow
and Frictional Resistance of Enclosed Rotating Disks. J. Basic Eng., voi. 82,
no. 1, Mar. 1960, pp. 217-232.
7. DAILY, J. W.; ERNST, W. D.; AND ASREDIAN, V. V.: Enclosed Rotating Disks with
Superposed Throughflow: Mean Study and Periodic Unsteady Characteristics
of the Induced Flow. Rep. R-$4-16, Massachusetts Inst. Tech. (AR01)-2500-2,
AD-443060), Apr. 1964.
8. STODOLA, A. (Louis C. LOEWENSTEIN, TRANS.): Steam and Gas Turbines. Vol. I.
McGraw-Hill Book Co., Inc., 1927. Reprinted by Peter Smith, 1945, pp. 200-201.
9. KLASSEN, HUGH A.: Cold-Air Investigation of Effects of Partial Admission on
Performance of 3.75-Inch Mean-Diameter Single-Stage Axial-Flow Turbine.
NASA TN D-4700, 1968.
10. STENNING, ALAN H.: Design of Turbines for High-Energy-Euel Low-Power-
Output Applications. Rep. 79, Dynamic Analysis and Control Lab., Massachu-
setts Inst. Tech., Sept. 30, 1953.
11. AINLEY, ]). G.; AND MATHIESON, G. C. R.: All Exambmtion of the Flow and
Pre_ure Losses in Blade Rows of Axial-Flow Turbines. R&M-2891, Aero-
nautical Research Council, Gt. Britain, 1955.
12. FLAGG, E. E.: Analytical Procedure and Computer Program for Determining the
Off-Design Performance of Axial-Flow Turbines. NASA CR-710, 1967.
146
MISCELLANEOUS LOSSES
A
a
Cxi
Clv
c:CM ,o
D
fgAh'
Ahid
i
J
Kd/
Ko
KpK,
K,_
Li1M
N
n
Pd/
Pp
PQR
r
8
U
V
W
Ol
P
p
SYMBOLS
area on one side of rotor disk, m_; ft _disk rim radius, m; ft
coefficient used to evaluate CM.o in regime II by equation (8-11)
coefficient used to evaluate CM.o in regime IV by equation (8-15)fluid shear-stress coefficient
torque coefficient with no throughflowdiameter, m; ft
nozzle active arc length, m; ft
convcrsion constant, l; 32.17 (lbm) (ft)/(lbf) (sec _)turbine specific work, J/kg; Btu/lb
turbine ideal specific work based on ratio of inlet-total pressureto exit-static pressure, J/kg; Btu/lb
incidence angle, deg
conversion constant, 1 ; 778 (ft) (lb)/Btu
disk-friction power-loss coefficient
ratio of rotating-core angular velocity to disk angular velocity
pumping power loss coefficient, 1/ml/2; (lbf)(sec2)/(lbm)(ft 3/2)sector loss coefficient
rotor velocity coefficient for full-admission impulse turbineincidence loss, J/kg; Btu/lb
blade height, m; ft
frictional resistance torque for both sides of rotor disk, N-m;lb-ft
rotative speed, rad/sec; rev/min
exponent in equation (8-35)
disk-friction power loss, W ; Btu/secpumping power loss, W ; Btu/sec
rotor-blade pitch, m; ft
volumetric throughflow rate, m'_/scc ; ft'_/secReynolds number
radius, m; ft
axial distance between rotor disk and casing, m; ftblade speed, m/sec; ft/sec
absolute velocity, m/sec ; ft/sec
relatiw, velocity, m/see; ft/sec
fluid flow angle from axial direction, deg
blade inlet angle from axial direction, deg
fluid relative angle measured from axial direction, degactive fraction of stator exit area
turbine static efficiency
dynamic viscosity, (N) (sec)/m2; lb/(ft) (sec)
density, kg/m._; lb/ft a
147
TURBINE DESIGN AND APPLICATION
fluid shear stress, N/m_; lb/ft 2 -_
T throughflow number defined by equation (8-21)angular velocity, rad/sec
Subscripts:
m mean section
n component normal to blade inlet dircction
o no throughflow
opt optimum
p component parallel to blade inlet direction
pa partial admissionr disk rim
u tangential component
_rotor inlet1 [blade-row inlet2 rotor exit
',_ U.S. GOVERNMENT PRINTING OFFICE: 1974. 0---,523-092
148